JP4921611B2 - Speech decoding apparatus, speech decoding method, and speech decoding program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To improve the subjective quality of a decoded signal by reducing generated pre-echoes and post-echoes without significantly increasing a bit rate in a band extension technology in a frequency region represented by SBR. <P>SOLUTION: A voice decoding device obtains a linear prediction coefficient by performing linear prediction analysis in a frequency direction to a signal represented by the frequency region based on a covariance method or an autocorrelation method, and deforms a time envelope of the signal by adjusting a filter intensity for the obtained linear prediction coefficient and then by filtering the signal in the frequency direction based on the adjusted coefficient. <P>COPYRIGHT: (C)2012,JPO&amp;INPIT

Description

  The present invention relates to a speech encoding device, speech decoding device, speech encoding method, speech decoding method, speech encoding program, and speech decoding program.

  Audio-acoustic coding technology that compresses the amount of signal data to several tenths by removing information unnecessary for human perception using auditory psychology is an extremely important technology for signal transmission and storage. . Examples of widely used perceptual audio coding techniques include “MPEG4 AAC” standardized by “ISO / IEC MPEG”.

  As a method for further improving the performance of speech coding and obtaining high speech quality at a low bit rate, a band expansion technique for generating a high-frequency component using a low-frequency component of speech has been widely used in recent years. A typical example of the bandwidth extension technology is SBR (SpectralBand Replication) technology used in “MPEG4 AAC”. In SBR, a signal converted into a frequency domain by a QMF (Quadrature MirrorFilter) filter bank is used to generate a high frequency component by copying a spectrum coefficient from a low frequency band to a high frequency band, and then the spectrum of the copied coefficient. The high frequency component is adjusted by adjusting the envelope and tonality. The speech coding method using the band expansion technology can reproduce the high-frequency component of the signal using only a small amount of auxiliary information, and is therefore effective for reducing the bit rate of speech coding.

  Band extension technology in the frequency domain typified by SBR is to adjust the spectral envelope and tonality for the spectral coefficients expressed in the frequency domain, adjust the gain for the spectral coefficients, linear prediction inverse filtering in the time direction, noise This is done by superimposing. With this adjustment process, when a signal having a large time envelope change such as a speech signal, applause, or castanets is encoded, reverberant noise called pre-echo or post-echo may be perceived in the decoded signal. This problem is caused by the time envelope of the high-frequency component being deformed during the adjustment process, and in many cases, the shape becomes flatter than before the adjustment. The time envelope of the high frequency component flattened by the adjustment processing does not coincide with the time envelope of the high frequency component in the original signal before the sign, and causes pre-echo and post-echo.

  Similar pre-echo and post-echo problems also occur in multi-channel acoustic coding using parametric processing, represented by “MPEG Surround” and parametric stereo. The decoder in multi-channel acoustic coding includes means for applying a decorrelation process to the decoded signal using a reverberation filter, but the time envelope of the signal is deformed in the process of the decorrelation process, and a reproduced signal similar to a pre-echo / post-echo signal Degradation occurs. As a solution to this problem, TES (Temporal Envelope Shaping) technology exists (Patent Document 1). In the TES technique, linear prediction analysis is performed in the frequency direction on a signal before decorrelation processing expressed in the QMF region, and after obtaining a linear prediction coefficient, the signal after decorrelation processing is performed using the obtained linear prediction coefficient. Is subjected to linear prediction synthesis filter processing in the frequency direction. With this process, the TES technique extracts the time envelope of the signal before the decorrelation process, and adjusts the time envelope of the signal after the decorrelation process accordingly. Since the signal before decorrelation processing has a time envelope with less distortion, the above processing adjusts the time envelope of the signal after decorrelation processing to a shape with less distortion, and improves pre-echo and post-echo reproduction. A signal can be obtained.

US Patent Application Publication No. 2006/0239473

  The TES technique described above utilizes the fact that a signal before decorrelation processing has a time envelope with little distortion. However, since the SBR decoder duplicates the high frequency component of the signal by copying the signal from the low frequency component, it is not possible to obtain a time envelope with little distortion related to the high frequency component. One solution to this problem is to analyze the high-frequency component of the input signal in the SBR encoder, quantize the linear prediction coefficient obtained as a result of the analysis, and multiplex it into a bitstream for transmission. As a result, a linear prediction coefficient including information with little distortion regarding the time envelope of the high frequency component can be obtained in the SBR decoder. However, in this case, a large amount of information is required for transmission of the quantized linear prediction coefficient, which causes a problem that the bit rate of the entire encoded bit stream is remarkably increased. Accordingly, an object of the present invention is to reduce the generated pre-echo and post-echo and improve the subjective quality of the decoded signal without significantly increasing the bit rate in the band expansion technology in the frequency domain represented by SBR. It is.

  The speech coding apparatus of the present invention is a speech coding apparatus that encodes a speech signal, and includes a core coding unit that encodes a low frequency component of the speech signal, and a time envelope of the low frequency component of the speech signal. Using time envelope auxiliary information calculating means for calculating time envelope auxiliary information for obtaining an approximation of the time envelope of the high frequency component of the audio signal, and at least the low frequency component encoded by the core encoding means And bit stream multiplexing means for generating a bit stream in which the time envelope auxiliary information calculated by the time envelope auxiliary information calculating means is multiplexed.

  In the speech coding apparatus according to the present invention, it is preferable that the time envelope auxiliary information represents a parameter indicating the steepness of change of the time envelope in the high frequency component of the speech signal within a predetermined analysis interval.

  The speech coding apparatus according to the present invention further comprises frequency conversion means for converting the speech signal into a frequency domain, wherein the time envelope auxiliary information calculation means is a high frequency of the speech signal converted into the frequency domain by the frequency conversion means. It is preferable to calculate the time envelope auxiliary information based on a high-frequency linear prediction coefficient obtained by performing a linear prediction analysis on the side coefficient in the frequency direction.

  In the speech coding apparatus of the present invention, the time envelope auxiliary information calculating means performs a linear prediction analysis in a frequency direction on a low frequency side coefficient of the speech signal converted into a frequency domain by the frequency converting means, and performs low frequency It is preferable to obtain a linear prediction coefficient and calculate the temporal envelope auxiliary information based on the low frequency linear prediction coefficient and the high frequency linear prediction coefficient.

  In the speech encoding device of the present invention, the temporal envelope auxiliary information calculating means acquires a prediction gain from each of the low-frequency linear prediction coefficient and the high-frequency linear prediction coefficient, and based on the magnitude of the two prediction gains, It is preferable to calculate time envelope auxiliary information.

  In the speech encoding device of the present invention, the time envelope auxiliary information calculating means separates a high frequency component from the speech signal, acquires time envelope information expressed in a time domain from the high frequency component, and It is preferable to calculate the time envelope auxiliary information based on the magnitude of the temporal change.

  In the speech coding apparatus of the present invention, the time envelope auxiliary information is obtained by using a low-frequency linear prediction coefficient obtained by performing a linear prediction analysis in a frequency direction on a low-frequency component of the speech signal. It is preferable to include difference information for acquisition.

  The speech coding apparatus according to the present invention further comprises frequency conversion means for converting the speech signal into a frequency domain, wherein the time envelope auxiliary information calculation means is a low-frequency unit for converting the speech signal converted into the frequency domain by the frequency conversion means. A linear prediction analysis is performed in the frequency direction for each of the frequency component and the high frequency side coefficient to obtain a low frequency linear prediction coefficient and a high frequency linear prediction coefficient, and a difference between the low frequency linear prediction coefficient and the high frequency linear prediction coefficient is obtained. It is preferable that the difference information is acquired.

  In the speech encoding device of the present invention, the difference information is linear prediction in any region of LSP (Linear Spectrum Pair), ISP (Immittance Spectrum Pair), LSF (Linear Spectrum Frequency), ISF (Immittance Spectrum Frequency), and PARCOR coefficient. It is preferable to represent the difference between the coefficients.

  The speech coding apparatus according to the present invention is a speech coding apparatus that encodes a speech signal, and includes a core coding unit that encodes a low frequency component of the speech signal, and a frequency that converts the speech signal into a frequency domain. Conversion means, linear prediction analysis means for obtaining a high-frequency linear prediction coefficient by performing linear prediction analysis in a frequency direction on the high-frequency side coefficient of the speech signal converted into the frequency domain by the frequency conversion means, and the linear prediction analysis Prediction coefficient thinning means for thinning out the high-frequency linear prediction coefficient acquired by the means in the time direction, prediction coefficient quantization means for quantizing the high-frequency linear prediction coefficient after thinning out by the prediction coefficient thinning means, The low frequency component after encoding by the core encoding means and the high frequency linear prediction coefficient after quantization by the prediction coefficient quantization means are multiplexed. Comprising a bit stream multiplexing means for generating a bit stream, and wherein the.

  A speech decoding apparatus according to the present invention is a speech decoding apparatus that decodes an encoded speech signal, wherein an external bit stream including the encoded speech signal is converted into an encoded bit stream, time envelope auxiliary information, and A bit stream separating means for separating the encoded bit stream, a core decoding means for decoding the encoded bit stream separated by the bit stream separating means to obtain a low frequency component, and a low frequency component obtained by the core decoding means. Frequency conversion means for converting to a frequency domain, high frequency generation means for generating a high frequency component by copying the low frequency component converted to the frequency domain by the frequency conversion means from a low frequency band to a high frequency band, and the frequency conversion Time envelope information by analyzing the low frequency component transformed into the frequency domain by means Low frequency time envelope analyzing means to be obtained, time envelope adjusting means for adjusting the time envelope information acquired by the low frequency time envelope analyzing means using the time envelope auxiliary information, and after adjustment by the time envelope adjusting means Time envelope deformation means for deforming the time envelope of the high-frequency component generated by the high-frequency generation means using the time envelope information.

  The speech decoding apparatus according to the present invention further includes a high frequency adjusting means for adjusting the high frequency component, and the frequency converting means is a 64-division QMF filter bank having real or complex coefficients, and the frequency converting means and the high frequency generating means. The high-frequency adjusting means preferably operates in accordance with an SBR decoder (SBR: SpectralBand Replication) in “MPEG4 AAC” defined in “ISO / IEC 14496-3”.

  In the speech decoding apparatus according to the present invention, the low frequency temporal envelope analysis means obtains a low frequency linear prediction coefficient by performing a linear prediction analysis in a frequency direction on the low frequency component converted into the frequency domain by the frequency conversion means. The time envelope adjusting means adjusts the low frequency linear prediction coefficient using the time envelope auxiliary information, and the time envelope deforming means applies the high frequency component of the frequency domain generated by the high frequency generating means to the high frequency component. It is preferable to perform linear prediction filter processing in the frequency direction using the linear prediction coefficient adjusted by the time envelope adjusting means to deform the time envelope of the audio signal.

  In the speech decoding apparatus of the present invention, the low frequency time envelope analyzing means obtains the power of each time slot of the low frequency component converted into the frequency domain by the frequency converting means, thereby obtaining time envelope information of the speech signal. The time envelope adjusting means adjusts the time envelope information using the time envelope auxiliary information, and the time envelope deforming means adds the high frequency component of the frequency domain generated by the high frequency generating means to the high frequency component after the adjustment. It is preferable to deform the time envelope of the high-frequency component by superimposing the time envelope information.

  In the speech decoding apparatus of the present invention, the low frequency time envelope analyzing means obtains the power for each QMF subband sample of the low frequency component converted into the frequency domain by the frequency converting means, thereby obtaining a time envelope of the speech signal. Information is acquired, the time envelope adjusting means adjusts the time envelope information using the time envelope auxiliary information, and the time envelope deforming means adds the high frequency component of the frequency domain generated by the high frequency generating means to the high frequency component. Preferably, the time envelope of the high frequency component is deformed by multiplying the adjusted time envelope information.

  In the speech decoding apparatus of the present invention, it is preferable that the temporal envelope auxiliary information represents a filter strength parameter for use in adjusting the strength of the linear prediction coefficient.

  In the speech decoding apparatus according to the present invention, it is preferable that the time envelope auxiliary information represents a parameter indicating a magnitude of a time change of the time envelope information.

  In the speech decoding apparatus of the present invention, it is preferable that the temporal envelope auxiliary information includes difference information of a linear prediction coefficient with respect to the low frequency linear prediction coefficient.

  In the speech decoding apparatus according to the present invention, the difference information is a linear prediction coefficient in any region of LSP (Linear Spectrum Pair), ISP (Immittance Spectrum Pair), LSF (Linear Spectrum Frequency), ISF (Immittance Spectrum Frequency), and PARCOR coefficient. It is preferable to express the difference between the two.

  In the speech decoding apparatus of the present invention, the low frequency temporal envelope analyzing means performs a linear prediction analysis in a frequency direction on the low frequency component converted into the frequency domain by the frequency converting means to obtain the low frequency linear prediction coefficient. And acquiring time envelope information of the audio signal by acquiring power for each time slot of the low frequency component in the frequency domain, and the time envelope adjusting means uses the time envelope auxiliary information to acquire the low frequency envelope information. Adjusting the frequency linear prediction coefficient and adjusting the time envelope information using the time envelope auxiliary information, and the time envelope deforming means adjusts the time envelope for the high frequency component of the frequency domain generated by the high frequency generating means. Frequency using linear prediction coefficients adjusted by the means A time envelope of the high frequency component by superimposing the time envelope information adjusted by the time envelope adjusting means on the high frequency component of the frequency domain by transforming the time envelope of the speech signal by performing a linear prediction filter processing of direction Is preferably modified.

  In the speech decoding apparatus of the present invention, the low frequency temporal envelope analyzing means performs a linear prediction analysis in a frequency direction on the low frequency component converted into the frequency domain by the frequency converting means to obtain the low frequency linear prediction coefficient. And acquiring time envelope information of the audio signal by acquiring power for each QMF subband sample of the low frequency component in the frequency domain, and the time envelope adjusting means uses the time envelope auxiliary information. The low-frequency linear prediction coefficient is adjusted and the time envelope information is adjusted using the time envelope auxiliary information, and the time envelope deforming means is configured to adjust the time for high frequency components in the frequency domain generated by the high frequency generating means. Use linear prediction coefficient after adjustment by envelope adjustment means The frequency envelope linear predictive filter processing is performed to transform the time envelope of the audio signal, and the high frequency component in the frequency domain is multiplied by the time envelope information adjusted by the time envelope adjusting means. It is preferable to deform the time envelope.

  In the speech decoding apparatus of the present invention, it is preferable that the time envelope auxiliary information represents a parameter indicating both the filter strength of the linear prediction coefficient and the time change magnitude of the time envelope information.

  A speech decoding apparatus according to the present invention is a speech decoding apparatus that decodes an encoded speech signal, and converts an external bit stream including the encoded speech signal into an encoded bit stream and a linear prediction coefficient. Bit stream separation means for separating, linear prediction coefficient interpolation / extrapolation means for interpolating or extrapolating the linear prediction coefficient in the time direction, and linear prediction coefficients interpolated or extrapolated by the linear prediction coefficient interpolation / extrapolation means And a time envelope deforming means for deforming the time envelope of the audio signal by performing linear prediction filter processing in the frequency direction on the high-frequency component expressed in the frequency domain.

  The speech encoding method of the present invention is a speech encoding method using a speech encoding device that encodes a speech signal, wherein the speech encoding device encodes a low-frequency component of the speech signal. And a time envelope assist in which the speech coding apparatus calculates time envelope assist information for obtaining an approximation of a time envelope of a high frequency component of the speech signal using a time envelope of a low frequency component of the speech signal. And at least the low-frequency component encoded in the core encoding step and the time envelope auxiliary information calculated in the time envelope auxiliary information calculation step are multiplexed by the speech encoding apparatus. And a bitstream multiplexing step for generating a bitstream.

  The speech encoding method of the present invention is a speech encoding method using a speech encoding device that encodes a speech signal, wherein the speech encoding device encodes a low-frequency component of the speech signal. Step, a frequency conversion step in which the speech encoding apparatus converts the speech signal into a frequency domain, and a high frequency side coefficient of the speech signal that the speech encoding apparatus has converted into the frequency domain in the frequency conversion step. A linear prediction analysis step for obtaining a high-frequency linear prediction coefficient by performing linear prediction analysis in the frequency direction, and a prediction coefficient by which the speech coding apparatus thins out the high-frequency linear prediction coefficient acquired in the linear prediction analysis step in the time direction. A thinning step; and the speech encoding apparatus quantizes the high-frequency linear prediction coefficient after the thinning in the prediction coefficient thinning step. And the speech encoding apparatus multiplexes at least the low frequency component after encoding in the core encoding step and the high frequency linear prediction coefficient after quantization in the prediction coefficient quantization step. And a bitstream multiplexing step for generating a generated bitstream.

  The speech decoding method of the present invention is a speech decoding method using a speech decoding device that decodes an encoded speech signal, and the speech decoding device includes an external bitstream including the encoded speech signal. A bit stream separating step for separating the encoded bit stream and the time envelope auxiliary information, and a core for obtaining a low frequency component by decoding the coded bit stream separated in the bit stream separating step by the speech decoding apparatus A decoding step; a frequency converting step in which the speech decoding apparatus converts the low-frequency component obtained in the core decoding step into a frequency domain; and the low-frequency component converted into the frequency domain in the frequency converting step. Generate high frequency components by copying frequency components from low frequency band to high frequency band A high frequency generation step, a low frequency time envelope analysis step in which the speech decoding apparatus analyzes the low frequency component converted into the frequency domain in the frequency conversion step to obtain time envelope information, and the speech decoding apparatus includes: A time envelope adjustment step of adjusting the time envelope information acquired in the low frequency time envelope analysis step using the time envelope auxiliary information; and the speech decoding apparatus adjusts the time envelope after the adjustment in the time envelope adjustment step. And a time envelope deformation step of deforming a time envelope of the high frequency component generated in the high frequency generation step using information.

  The speech decoding method of the present invention is a speech decoding method using a speech decoding device that decodes an encoded speech signal, and the speech decoding device includes an external bitstream including the encoded speech signal. A bit stream separation step of separating the encoded bit stream and linear prediction coefficients, a linear prediction coefficient interpolation / extrapolation step in which the speech decoding apparatus interpolates or extrapolates the linear prediction coefficients in the time direction, The speech decoding apparatus performs linear prediction filter processing in the frequency direction on the high-frequency component expressed in the frequency domain using the linear prediction coefficient interpolated or extrapolated in the linear prediction coefficient interpolation / extrapolation step to generate a speech signal A time envelope deformation step for deforming the time envelope.

  The speech encoding program of the present invention uses a computer device, core encoding means for encoding a low frequency component of the speech signal, and a time envelope of the low frequency component of the speech signal to encode the speech signal. Time envelope auxiliary information calculating means for calculating time envelope auxiliary information for obtaining an approximation of the time envelope of the high frequency component of the audio signal, and at least the low frequency component encoded by the core encoding means And a bit stream multiplexing means for generating a bit stream multiplexed with the time envelope auxiliary information calculated by the time envelope auxiliary information calculating means.

  In order to encode a speech signal, the speech encoding program of the present invention includes a computer device, a core encoding unit that encodes a low frequency component of the speech signal, and a frequency conversion unit that converts the speech signal into a frequency domain. , Linear prediction analysis means for obtaining a high-frequency linear prediction coefficient by performing linear prediction analysis in a frequency direction on the high-frequency side coefficient of the speech signal converted into the frequency domain by the frequency conversion means, and acquired by the linear prediction analysis means Further, prediction coefficient thinning means for thinning out the high-frequency linear prediction coefficient in the time direction, prediction coefficient quantization means for quantizing the high-frequency linear prediction coefficient after thinning out by the prediction coefficient thinning-out means, and at least the core coding means The low-frequency component after encoding by the high-frequency linear prediction coefficient after quantization by the prediction coefficient quantization means, Characterized in that to function bit stream multiplexing means for generating the bitstream as.

  In order to decode an encoded audio signal, the audio decoding program of the present invention uses a computer device to convert an external bit stream including the encoded audio signal into an encoded bit stream, time envelope auxiliary information, A bit stream separating means for separating the encoded bit stream by the bit stream separating means to obtain a low frequency component, and a low frequency component obtained by the core decoding means in the frequency domain A frequency converting means for converting to a frequency region, a high frequency generating means for generating a high frequency component by copying the low frequency component converted into the frequency region by the frequency converting means from a low frequency band to a high frequency band, Analyzing the low-frequency component converted into a time envelope information By the low-frequency time envelope analyzing means for acquiring the time envelope information acquired by the low-frequency time envelope analyzing means by using the time envelope auxiliary information, and the time envelope adjusting means. The time envelope information after adjustment is used to function as time envelope deformation means for deforming the time envelope of the high frequency component generated by the high frequency generation means.

  In order to decode an encoded audio signal, an audio decoding program according to the present invention converts a computer apparatus into an external bit stream including the encoded audio signal into an encoded bit stream and a linear prediction coefficient. Bit stream separation means for separating, linear prediction coefficient interpolation / extrapolation means for interpolating or extrapolating the linear prediction coefficient in the time direction, and linear prediction coefficients interpolated or extrapolated by the linear prediction coefficient interpolation / extrapolation means And a high-frequency component expressed in the frequency domain by performing linear prediction filter processing in the frequency direction to function as time envelope deformation means for deforming the time envelope of the audio signal.

  In the speech decoding apparatus of the present invention, the time envelope deforming unit performs a linear prediction filter process in the frequency direction on the high frequency component in the frequency domain generated by the high frequency generating unit, and then results of the linear prediction filter process. It is preferable to adjust the power of the obtained high frequency component to a value equal to that before the linear prediction filter processing.

  In the speech decoding apparatus of the present invention, the time envelope deforming unit performs a linear prediction filter process in the frequency direction on the high frequency component in the frequency domain generated by the high frequency generating unit, and then results of the linear prediction filter process. It is preferable to adjust the power within an arbitrary frequency range of the obtained high frequency component to a value equal to that before the linear prediction filter processing.

  In the speech decoding apparatus according to the present invention, it is preferable that the time envelope auxiliary information is a ratio between a minimum value and an average value in the adjusted time envelope information.

  In the speech decoding apparatus according to the present invention, the time envelope deforming means may adjust the time envelope after the adjustment so that the power in the SBR envelope time segment of the high frequency component in the frequency domain becomes equal before and after the deformation of the time envelope. After controlling the gain, it is preferable to transform the time envelope of the high frequency component by multiplying the high frequency component of the frequency domain by the gain-controlled time envelope.

  In the speech decoding apparatus of the present invention, the low frequency time envelope analyzing means acquires power for each QMF subband sample of the low frequency component converted into the frequency domain by the frequency converting means, and further, within the SBR envelope time segment. It is preferable to obtain time envelope information expressed as a gain coefficient to be multiplied to each QMF subband sample by normalizing the power for each QMF subband sample using the average power at.

  The speech decoding apparatus according to the present invention is a speech decoding apparatus that decodes an encoded speech signal, and that decodes an external bit stream including the encoded speech signal to obtain a low frequency component. And a frequency converting means for converting the low frequency component obtained by the core decoding means to a frequency domain, and copying the low frequency component converted to the frequency domain by the frequency converting means from a low frequency band to a high frequency band. A high-frequency generating means for generating a high-frequency component, a low-frequency time envelope analyzing means for analyzing the low-frequency component converted into the frequency domain by the frequency converting means to obtain time envelope information, and analyzing the bitstream A time envelope auxiliary information generating unit for generating time envelope auxiliary information, and the low frequency time envelope The time envelope information acquired by the analysis means is adjusted using the time envelope auxiliary information, and the time envelope information adjusted by the time envelope adjustment means is used by the high frequency generation means. And a time envelope deforming means for deforming the generated time envelope of the high-frequency component.

  The speech decoding apparatus of the present invention includes a primary high-frequency adjusting unit and a secondary high-frequency adjusting unit corresponding to the high-frequency adjusting unit, and the primary high-frequency adjusting unit is a part of the process corresponding to the high-frequency adjusting unit. The time envelope deformation means performs time envelope deformation on the output signal of the primary high frequency adjustment means, and the secondary high frequency adjustment means applies to the output signal of the time envelope deformation means. Of the processes corresponding to the high-frequency adjusting means, it is preferable to execute a process that is not executed by the primary high-frequency adjusting means, and the secondary high-frequency adjusting means is preferably a sine wave addition process in the SBR decoding process. .

  According to the present invention, it is possible to reduce the generated pre-echo and post-echo and improve the subjective quality of the decoded signal without significantly increasing the bit rate in the band expansion technique in the frequency domain represented by SBR.

It is a figure which shows the structure of the audio | voice coding apparatus which concerns on 1st Embodiment. It is a flowchart for demonstrating operation | movement of the audio | voice coding apparatus which concerns on 1st Embodiment. It is a figure which shows the structure of the audio | voice decoding apparatus which concerns on 1st Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on 1st Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the modification 1 of 1st Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on 2nd Embodiment. It is a flowchart for demonstrating operation | movement of the audio | voice coding apparatus which concerns on 2nd Embodiment. It is a figure which shows the structure of the audio | voice decoding apparatus which concerns on 2nd Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on 2nd Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on 3rd Embodiment. It is a flowchart for demonstrating operation | movement of the audio | voice coding apparatus which concerns on 3rd Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on 3rd Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on 3rd Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the audio | voice decoding apparatus which concerns on the other modification of 1st Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 1st Embodiment. It is a figure which shows the structure of the audio | voice decoding apparatus which concerns on the other modification of 1st Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 1st Embodiment. It is a figure which shows the structure of the audio | voice decoding apparatus which concerns on the modification of 2nd Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the modification of 2nd Embodiment. It is a figure which shows the structure of the audio | voice decoding apparatus which concerns on the other modification of 2nd Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 2nd Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a flowchart for demonstrating operation | movement of the speech decoding apparatus which concerns on the other modification of 4th Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the other modification of 1st Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the other modification of 1st Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the modification of 2nd Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the other modification of 2nd Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on 4th Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the modification of 4th Embodiment. It is a figure which shows the structure of the audio | voice coding apparatus which concerns on the other modification of 4th Embodiment.

  Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings. In the description of the drawings, if possible, the same elements are denoted by the same reference numerals, and redundant description is omitted.

(First embodiment)
FIG. 1 is a diagram illustrating a configuration of a speech encoding device 11 according to the first embodiment. The speech encoding device 11 is physically provided with a CPU, ROM, RAM, communication device, etc. (not shown), and this CPU is a predetermined computer program (for example, stored in the internal memory of the speech encoding device 11 such as a ROM). The computer program for executing the processing shown in the flowchart of FIG. The communication device of the audio encoding device 11 receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside.

  The speech encoding device 11 functionally includes a frequency converting unit 1a (frequency converting unit), a frequency inverse converting unit 1b, a core codec encoding unit 1c (core encoding unit), an SBR encoding unit 1d, and a linear prediction analysis. Unit 1e (time envelope auxiliary information calculating unit), filter strength parameter calculating unit 1f (time envelope auxiliary information calculating unit), and bitstream multiplexing unit 1g (bitstream multiplexing unit). The frequency conversion unit 1a to the bit stream multiplexing unit 1g of the speech encoding device 11 shown in FIG. 1 are executed by the CPU of the speech encoding device 11 executing a computer program stored in the built-in memory of the speech encoding device 11. This is a function that is realized. The CPU of the speech encoding device 11 executes the computer program (using the frequency converting unit 1a to the bitstream multiplexing unit 1g shown in FIG. 1), thereby performing the processing (step Sa1 to step 1) shown in the flowchart of FIG. The process of Sa7) is executed sequentially. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech encoding device 11.

  The frequency converting unit 1a analyzes the input signal received from the outside via the communication device of the speech encoding device 11 using the multi-divided QMF filter bank, and obtains a signal q (k, r) in the QMF region (step Sa1). Processing). Here, k (0 ≦ k ≦ 63) is an index in the frequency direction, and r is an index indicating a time slot. The frequency inverse transform unit 1b synthesizes half of the low frequency side coefficients of the signal in the QMF region obtained from the frequency transform unit 1a by the QMF filter bank, and is downsampled including only the low frequency component of the input signal. A time domain signal is obtained (processing of step Sa2). The core codec encoding unit 1c encodes the down-sampled time domain signal to obtain an encoded bit stream (processing of step Sa3). The encoding in the core codec encoding unit 1c may be based on a speech encoding method typified by the CELP method, or based on acoustic coding such as transform coding typified by AAC or TCX (Transform Coded Excitation) method. May be.

  The SBR encoding unit 1d receives the signal in the QMF region from the frequency conversion unit 1a, performs SBR encoding based on the analysis of the power, signal change, tonality, etc. of the high frequency component to obtain SBR auxiliary information (processing of step Sa4) ). The QMF analysis method in the frequency conversion unit 1a and the SBR encoding method in the SBR encoding unit 1d are described in detail in, for example, the document “3GPP TS 26.404; Enhanced aacPlus encoder SBR part”.

The linear prediction analysis unit 1e receives a signal in the QMF region from the frequency conversion unit 1a, performs linear prediction analysis on the high frequency component of this signal in the frequency direction, and performs a high frequency linear prediction coefficient a _H (n, r) (1 ≦ n). ≦ N) is acquired (processing of step Sa5). N is the linear prediction order. The index r is an index in the time direction regarding the subsample of the signal in the QMF region. A covariance method or an autocorrelation method can be used for signal linear prediction analysis. The linear prediction analysis for acquiring a _H (n, r) is performed on the high frequency component satisfying k _x <k ≦ 63 in q (k, r). However k _x is a frequency index corresponding to the upper limit frequency of the frequency band to be encoded by the core codec encoding unit 1c. Also, the linear prediction analysis unit 1e performs linear predictive analysis on another low-frequency component that was analyzed in obtaining a H _(n, r), different from the a H _(n, r) Of the low frequency linear prediction coefficient a _L (n, r) may be acquired (the linear prediction coefficient related to such a low frequency component corresponds to the time envelope information, and in the first embodiment, The same). The linear prediction analysis when acquiring a _L (n, r) is for low frequency components satisfying 0 ≦ k <k _x . Further, this linear prediction analysis may be performed for a part of frequency bands included in a section of 0 ≦ k <k _x .

The filter strength parameter calculation unit 1f uses, for example, the linear prediction coefficient acquired by the linear prediction analysis unit 1e, and the filter strength parameter (the filter strength parameter corresponds to the time envelope auxiliary information. The same applies to step S6) (processing of step Sa6). First, the prediction gain G _H (r) is calculated from a _H (n, r). The calculation method of the prediction gain is described in detail, for example, in “Voice coding, Takehiro Moriya, edited by the Institute of Electronics, Information and Communication Engineers”. Further, when a _L (n, r) is calculated, the prediction gain G _L (r) is calculated in the same manner. The filter strength parameter K (r) is a parameter that increases as G _H (r) increases. For example, the filter strength parameter K (r) can be obtained according to the following mathematical formula (1). However, max (a, b) represents the maximum value of a and b, and min (a, b) represents the minimum value of a and b.

When G _L (r) is calculated, K (r) can be acquired as a parameter that increases as G _H (r) increases and decreases as G _L (r) increases. In this case, K can be obtained, for example, according to the following formula (2).

  K (r) is a parameter indicating the strength for adjusting the time envelope of the high frequency component during SBR decoding. The prediction gain for the linear prediction coefficient in the frequency direction increases as the time envelope of the signal in the analysis section shows a sharp change. K (r) is a parameter for instructing the decoder to increase the processing to sharpen the change in the time envelope of the high-frequency component generated by the SBR as the value increases. Note that K (r) is a parameter for instructing the decoder (for example, the speech decoding device 21) to weaken the processing for sharpening the time envelope of the high-frequency component generated by the SBR as the value thereof is smaller. It may be included, and may include a value indicating that the process of making the time envelope steep is not executed. Further, K (r) representing a plurality of time slots may be transmitted without transmitting K (r) of each time slot. In order to determine a time slot interval sharing the same value of K (r), it is desirable to use SBR envelope time border information included in the SBR auxiliary information.

K (r) is quantized and then transmitted to the bitstream multiplexing unit 1g. It is desirable to calculate a representative K (r) for a plurality of time slots, for example by averaging K (r) for a plurality of time slots r prior to quantization. In addition, when transmitting K (r) representing a plurality of time slots, the calculation of K (r) is not performed independently from the result of analyzing each time slot as in Equation (2). K (r) representing them may be acquired from the analysis result of the entire section composed of a plurality of time slots. In this case, the calculation of K (r) can be performed, for example, according to the following formula (3). Here, mean (•) represents an average value in the section of the time slot represented by K (r).

  When transmitting K (r), it may be transmitted exclusively with the inverse filter mode information included in the SBR auxiliary information described in “ISO / IEC 14496-3 subpart 4 General Audio Coding”. That is, K (r) is not transmitted for the time slot for transmitting the inverse filter mode information of the SBR auxiliary information, and the inverse filter mode information of the SBR auxiliary information (for the time slot for transmitting K (r) ( Bs_invf_mode) in “ISO / IEC 14496-3subpart 4 General Audio Coding” may not be transmitted. In addition, you may add the information which shows which of the reverse filter mode information contained in K (r) or SBR auxiliary information is transmitted. Alternatively, K (r) and the inverse filter mode information included in the SBR auxiliary information may be combined and handled as one vector information, and this vector may be entropy encoded. At this time, a restriction may be applied to a combination of values of K (r) and the inverse filter mode information included in the SBR auxiliary information.

  The bitstream multiplexing unit 1g includes the encoded bitstream calculated by the core codec encoding unit 1c, the SBR auxiliary information calculated by the SBR encoding unit 1d, and the K ( r) are multiplexed, and a multiplexed bit stream (encoded multiplexed bit stream) is output via the communication device of the audio encoding device 11 (processing of step Sa7).

  FIG. 3 is a diagram illustrating the configuration of the speech decoding apparatus 21 according to the first embodiment. The speech decoding device 21 is physically provided with a CPU, ROM, RAM, communication device, etc. (not shown), and this CPU is a predetermined computer program (for example, a diagram) stored in a built-in memory of the speech decoding device 21 such as a ROM. 4 is loaded into the RAM and executed, whereby the speech decoding apparatus 21 is comprehensively controlled. The communication device of the speech decoding device 21 includes encoded multiplexed bits output from the speech encoding device 11, the speech encoding device 11a of Modification 1 described later, or the speech encoding apparatus of Modification 2 described later. The stream is received, and the decoded audio signal is output to the outside. As shown in FIG. 3, the audio decoding device 21 functionally includes a bit stream separation unit 2a (bit stream separation unit), a core codec decoding unit 2b (core decoding unit), and a frequency conversion unit 2c (frequency conversion unit). , Low frequency linear prediction analysis unit 2d (low frequency time envelope analysis unit), signal change detection unit 2e, filter strength adjustment unit 2f (time envelope adjustment unit), high frequency generation unit 2g (high frequency generation unit), high frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, a high frequency adjustment unit 2j (high frequency adjustment unit), a linear prediction filter unit 2k (time envelope transformation unit), a coefficient addition unit 2m, and a frequency inverse conversion unit 2n. The bit stream separation unit 2a to the inverse frequency conversion unit 2n of the speech decoding device 21 shown in FIG. 3 are realized by the CPU of the speech decoding device 21 executing a computer program stored in the internal memory of the speech decoding device 21. It is a function. The CPU of the speech decoding apparatus 21 executes the computer program (using the bit stream separation unit 2a to the envelope shape parameter calculation unit 1n shown in FIG. 3), thereby performing the processing shown in the flowchart of FIG. Step Sb11) is sequentially executed. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech decoding device 21.

  The bit stream separation unit 2a separates the multiplexed bit stream input via the communication device of the audio decoding device 21 into a filter strength parameter, SBR auxiliary information, and an encoded bit stream. The core codec decoding unit 2b decodes the encoded bitstream given from the bitstream separation unit 2a, and obtains a decoded signal including only the low frequency component (processing in step Sb1). At this time, the decoding method may be based on a speech encoding method typified by the CELP method, or may be based on acoustic encoding such as AAC or TCX (Transform Coded Excitation) method.

The frequency conversion unit 2c analyzes the decoded signal given from the core codec decoding unit 2b using a multi-division QMF filter bank, and obtains a signal q _dec (k, r) in the QMF region (processing in step Sb2). Here, k (0 ≦ k ≦ 63) is an index in the frequency direction, and r is an index indicating an index in the time direction regarding a subsample of a signal in the QMF domain.

The low frequency linear prediction analysis unit 2d performs linear prediction analysis on the q _dec (k, r) obtained from the frequency conversion unit 2c in the frequency direction with respect to each of the time slots r, and the low frequency linear prediction coefficient a _dec (n, r). ) Is acquired (processing of step Sb3). The linear prediction analysis is performed on the range of 0 ≦ k <k _x corresponding to the signal band of the decoded signal obtained from the core codec decoding unit 2b. Further, this linear prediction analysis may be performed for a part of frequency bands included in a section of 0 ≦ k <k _x .

２．ｐ（ｒ）を平滑化したエンベロープｐ_ｅｎｖ（ｒ）を次の数式（５）によって取得する。ただしαは０＜α＜１を満たす定数である。

３．ｐ（ｒ）とｐ_ｅｎｖ（ｒ）とを用いてＴ（ｒ）を次の数式（６）に従って取得する。ただしβは定数である。

以上に示した方法は電力の変化に基づく信号変化検出の単純な例であり、他のもっと洗練された方法により信号変化検出を行ってもよい。また、信号変化検出部２ｅは省略してもよい。 The signal change detection unit 2e detects a time change of the signal in the QMF region obtained from the frequency conversion unit 2c, and outputs it as a detection result T (r). The signal change can be detected by, for example, the following method.
1. The short-time power p (r) of the signal in the time slot r is obtained by the following equation (4).

2. An envelope p _env (r) obtained by smoothing p (r) is obtained by the following equation (5). However, α is a constant that satisfies 0 <α <1.

3. T (r) is obtained according to the following equation (6) using p (r) and p _env (r). Where β is a constant.

The method described above is a simple example of signal change detection based on power change, and signal change detection may be performed by another more sophisticated method. Further, the signal change detection unit 2e may be omitted.

さらに、信号変化検出部２ｅの出力Ｔ（ｒ）が得られる場合には、強度の調整は次の数式（８）に従って行ってもよい。

The filter strength adjustment unit 2f adjusts the filter strength with respect to a _dec (n, r) obtained from the low frequency linear prediction analysis unit 2d to obtain an adjusted linear prediction coefficient a _adj (n, r). (Process of step Sb4). The adjustment of the filter strength can be performed, for example, according to the following formula (7) using the filter strength parameter K received via the bit stream separation unit 2a.

Further, when the output T (r) of the signal change detection unit 2e is obtained, the intensity may be adjusted according to the following formula (8).

The high frequency generator 2g copies the QMF domain signal obtained from the frequency converter 2c from the low frequency band to the high frequency band, and generates a QMF domain signal q _exp (k, r) of the high frequency component (in step Sb5). processing). High-frequency generation is performed according to the HF generation method in the SBR of “MPEG4 AAC” (“ISO / IEC 14496-3 subpart 4 General Audio Coding”).

The high-frequency linear prediction analysis unit 2h performs a linear prediction analysis in the frequency direction for q _exp (k, r) generated by the high-frequency generation unit 2g, and calculates a high-frequency linear prediction coefficient a _exp (n, r). Obtain (process of step Sb6). The linear prediction analysis is performed on a range of k _x ≦ k ≦ 63 corresponding to the high frequency component generated by the high frequency generation unit 2g.

この線形予測逆フィルタ処理は、低周波側の係数から高周波側の係数に向かって行われてもよいし、その逆でもよい。線形予測逆フィルタ処理は、後段において時間エンベロープ変形を行う前に高周波成分の時間エンベロープを一旦平坦化しておくための処理であり、線形予測逆フィルタ部２ｉは省略されてもよい。また、高周波生成部２ｇからの出力に対して高周波成分への線形予測分析と逆フィルタ処理を行うかわりに、後述する高周波調整部２ｊからの出力に対して高周波線形予測分析部２ｈによる線形予測分析と線形予測逆フィルタ部２ｉによる逆フィルタ処理とを行ってもよい。さらに、線形予測逆フィルタ処理に用いる線形予測係数は、ａ_ｅｘｐ（ｎ，ｒ）ではなく、ａ_ｄｅｃ（ｎ，ｒ）又はａ_ａｄｊ（ｎ，ｒ）であってもよい。また、線形予測逆フィルタ処理に用いられる線形予測係数は、ａ_ｅｘｐ（ｎ，ｒ）に対してフィルタ強度調整を行って取得される線形予測係数ａ_{ｅｘｐ，ａｄｊ}（ｎ，ｒ）であってもよい。強度調整は、ａ_ａｄｊ（ｎ，ｒ）を取得する際と同様、例えば、次の数式（１０）に従って行われる。

The linear prediction inverse filter unit 2i performs a linear prediction inverse filter process on the signal in the high frequency band QMF region generated by the high frequency generation unit 2g and using a _exp (n, r) as a coefficient in the frequency direction (step) Processing of Sb7). The transfer function of the linear prediction inverse filter is as shown in the following equation (9).

This linear prediction inverse filter processing may be performed from the low frequency side coefficient to the high frequency side coefficient, or vice versa. The linear prediction inverse filter process is a process for once flattening the time envelope of the high frequency component before performing the time envelope deformation in the subsequent stage, and the linear prediction inverse filter unit 2i may be omitted. Further, instead of performing linear prediction analysis and inverse filter processing on the high frequency components for the output from the high frequency generation unit 2g, linear prediction analysis by the high frequency linear prediction analysis unit 2h is performed on the output from the high frequency adjustment unit 2j described later. And inverse filter processing by the linear prediction inverse filter unit 2i may be performed. Furthermore, the linear prediction coefficient used for the linear prediction inverse filter processing may be a _dec (n, r) or a _adj (n, r) instead of a _exp (n, r). Also, the linear prediction coefficients used for a linear prediction inverse _{filtering, a} exp _(n, r) linear prediction coefficient is obtained by performing a filtering strength adjustment to _{a exp,} even adj (n, r) Good. The intensity adjustment is performed according to the following formula (10), for example, as in the case of acquiring a _adj (n, r).

  The high frequency adjustment unit 2j adjusts the frequency characteristic and tonality of the high frequency component for the output from the linear prediction inverse filter unit 2i (processing in step Sb8). This adjustment is performed according to the SBR auxiliary information given from the bitstream separation unit 2a. The processing by the high frequency adjustment unit 2j is performed in accordance with the “HF adjustment” step in the SBR of “MPEG4 AAC”. For the signal in the QMF region of the high frequency band, linear prediction inverse filter processing in the time direction, gain adjustment and This adjustment is performed by superimposing noise. Details of the processing in the above steps are described in detail in “ISO / IEC 14496-3subpart 4 General Audio Coding”. As described above, the frequency conversion unit 2c, the high frequency generation unit 2g, and the high frequency adjustment unit 2j all operate in accordance with the SBR decoder in “MPEG4 AAC” defined in “ISO / IEC 14496-3”. To do.

この線形予測合成フィルタ処理によって、線形予測フィルタ部２ｋは、ＳＢＲに基づいて生成された高周波成分の時間エンベロープを変形する。 Linear prediction filter unit 2k high-frequency components _q adj (n, r) of the QMF domain signal outputted from the high frequency adjusting unit 2j to, using a filter strength adjusting unit 2f _a obtained from adj (n, r) Then, linear prediction synthesis filter processing is performed in the frequency direction (processing of step Sb9). The transfer function in the linear prediction synthesis filter processing is as shown in the following formula (11).

By this linear prediction synthesis filter processing, the linear prediction filter unit 2k deforms the time envelope of the high-frequency component generated based on SBR.

  The coefficient adding unit 2m adds the signal in the QMF region including the low frequency component output from the frequency conversion unit 2c and the signal in the QMF region including the high frequency component output from the linear prediction filter unit 2k, and adds the low frequency component. And a signal in the QMF region including both the high-frequency component (processing in step Sb10).

  The frequency inverse transformation unit 2n processes the signal in the QMF region obtained from the coefficient addition unit 2m by the QMF synthesis filter bank. As a result, a time-domain decoded speech signal including both the low-frequency component obtained by decoding of the core codec and the high-frequency component generated by SBR and whose time envelope is deformed by the linear prediction filter is obtained and obtained. The voice signal thus output is output to the outside via the built-in communication device (step Sb11 processing). In addition, the frequency inverse transform unit 2n transmits K (r) and K (r) exclusively when the inverse filter mode information of the SBR auxiliary information described in “ISO / IEC 14496-3subpart 4 General Audio Coding” is transmitted. ) Is transmitted and the inverse filter mode information of the SBR auxiliary information is not transmitted, using the inverse filter mode information of the SBR auxiliary information for at least one of the time slots before and after the time slot. The inverse filter mode information of the SBR auxiliary information of the time slot may be generated, or the inverse filter mode information of the SBR auxiliary information of the time slot may be set to a predetermined mode. On the other hand, for the time slot in which the inverse filter data of the SBR auxiliary information is transmitted and K (r) is not transmitted, the frequency inverse transform unit 2n applies to at least one time slot before and after the time slot. Using K (r), K (r) for the time slot may be generated, and K (r) for the time slot may be set to a predetermined value. Note that the frequency inverse transform unit 2n determines whether the transmitted information is K (r) or SBR auxiliary information based on information indicating whether K (r) or the inverse filter mode information of the SBR auxiliary information is transmitted. It may be determined whether it is reverse filter mode information.

(Modification 1 of the first embodiment)
FIG. 5 is a diagram illustrating a configuration of a modified example (speech encoding apparatus 11a) of the speech encoding apparatus according to the first embodiment. The speech encoding device 11a physically includes a CPU, ROM, RAM, and a communication device (not shown). This CPU stores a predetermined computer program stored in the internal memory of the speech encoding device 11a such as a ROM as a RAM. The voice encoding device 11a is comprehensively controlled by loading and executing. The communication device of the audio encoding device 11a receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside.

  As shown in FIG. 5, the speech encoding device 11a functionally replaces the linear prediction analysis unit 1e, the filter strength parameter calculation unit 1f, and the bit stream multiplexing unit 1g of the speech encoding device 11 with a high frequency frequency. An inverse conversion unit 1h, a short-time power calculation unit 1i (time envelope auxiliary information calculation unit), a filter strength parameter calculation unit 1f1 (time envelope auxiliary information calculation unit), and a bit stream multiplexing unit 1g1 (bit stream multiplexing unit) are provided. . The bit stream multiplexing unit 1g1 has the same function as the bit stream multiplexing unit 1g. The frequency conversion unit 1a to SBR encoding unit 1d, the high frequency inverse frequency conversion unit 1h, the short time power calculation unit 1i, the filter strength parameter calculation unit 1f1, and the bit stream multiplexing unit 1g1 of the speech encoding device 11a shown in FIG. This is a function realized by the CPU of the speech encoding device 11a executing a computer program stored in the built-in memory of the speech encoding device 11a. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech encoding device 11a.

The high frequency inverse frequency transform unit 1h replaces the coefficient corresponding to the low frequency component encoded by the core codec encoding unit 1c among the signals in the QMF region obtained from the frequency conversion unit 1a with “0”, and then performs QMF. Processing is performed using the synthesis filter bank to obtain a time-domain signal including only high-frequency components. The short-time power calculation unit 1i calculates the power by dividing the time-domain high-frequency component obtained from the high-frequency inverse frequency conversion unit 1h into short sections, and calculates p (r). As an alternative method, the short-time power may be calculated according to the following equation (12) using a signal in the QMF region.

The filter strength parameter calculation unit 1f1 detects a change portion of p (r), and determines the value of K (r) so that K (r) increases as the change increases. The value of K (r) may be performed, for example, by the same method as the calculation of T (r) in the signal change detection unit 2e of the speech decoding device 21. Further, signal change detection may be performed by other more sophisticated methods. Further, the filter strength parameter calculation unit 1f1 acquires the low frequency by the same method as the calculation of T (r) in the signal change detection unit 2e of the speech decoding apparatus 21 after acquiring the power for a short time for each of the low frequency component and the high frequency component. The signal changes Tr (r) and Th (r) of each of the component and the high frequency component may be acquired and the value of K (r) may be determined using these. In this case, K (r) can be obtained, for example, according to the following formula (13). However, ε is a constant such as 3.0.

(Modification 2 of the first embodiment)
A speech encoding apparatus (not shown) of Modification 2 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically shown, and this CPU is a speech of Modification 2 such as a ROM. A predetermined computer program stored in the internal memory of the encoding device is loaded into the RAM and executed, whereby the speech encoding device according to the second modification is comprehensively controlled. The communication device of the audio encoding device of Modification 2 receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside.

  Functionally, the speech coding apparatus according to the second modified example is replaced with a linear prediction coefficient difference coding unit (time envelope) (not shown) instead of the filter strength parameter calculation unit 1f and the bitstream multiplexing unit 1g of the speech coding device 11. Auxiliary information calculating means) and a bit stream multiplexing section (bit stream multiplexing means) for receiving the output from the linear prediction coefficient difference encoding section. The frequency conversion unit 1a to the linear prediction analysis unit 1e, the linear prediction coefficient difference encoding unit, and the bitstream multiplexing unit of the speech encoding device of Modification 2 are modified by the CPU of the speech encoding device of Modification 2. This is a function realized by executing a computer program stored in the built-in memory of the second speech encoding apparatus. Various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech encoding apparatus according to the second modification. To do.

The linear prediction coefficient difference encoding unit uses the input signal a _H (n, r) and the input signal a _L (n, r), and uses the linear prediction coefficient difference value a _D (n) according to the following equation (14). , R).

The linear prediction coefficient difference encoding unit further quantizes a _D (n, r) and transmits the quantized bit to the bit stream multiplexing unit (configuration corresponding to the bit stream multiplexing unit 1g). The bit stream multiplexing unit multiplexes a _D (n, r) instead of K (r) into the bit stream, and outputs the multiplexed bit stream to the outside via a communication device incorporating the multiplexed bit stream.

  The speech decoding apparatus (not shown) of Modification 2 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not physically shown, and this CPU is a speech decoding of Modification 2 of the ROM or the like. A predetermined computer program stored in the built-in memory of the apparatus is loaded into the RAM and executed, whereby the speech decoding apparatus of the modified example 2 is comprehensively controlled. The communication device of the speech decoding apparatus according to the second modification includes the encoded speech output from the speech encoding apparatus 11, the speech encoding apparatus 11a according to the first modification, or the speech encoding apparatus according to the second modification. The bit stream is received, and the decoded audio signal is output to the outside.

  Functionally, the speech decoding apparatus according to the second modification includes a linear prediction coefficient difference decoding unit (not shown) instead of the filter strength adjustment unit 2f of the speech decoding device 21. The bit stream separation unit 2a to the signal change detection unit 2e, the linear prediction coefficient difference decoding unit, and the high frequency generation unit 2g to the frequency inverse conversion unit 2n of the speech decoding device according to the second modification are the CPUs of the speech decoding device according to the second modification. Is a function realized by executing a computer program stored in the internal memory of the speech decoding apparatus according to the second modification. Various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech decoding apparatus according to the second modification. .

The linear prediction coefficient difference decoding unit uses a _L (n, r) obtained from the low-frequency linear prediction analysis unit 2d and a _D (n, r) given from the bitstream separation unit 2a to obtain the following formula: According to (15), a _adj (n, r) subjected to differential decoding is obtained.

The linear prediction coefficient differential decoding unit transmits a _adj (n, r) differentially decoded in this way to the linear prediction filter unit 2k. a _D (n, r) may be a difference value in the prediction coefficient region as shown in Equation (14), but the prediction coefficient is represented by LSP (Linear Spectrum Pair), ISP (ImmittanceSpectrum Pair), LSF ( It may be a value obtained by taking a difference after conversion to another expression format such as Linear Spectrum Frequency (ISF), ISF (Immittance Spectrum Frequency), or PARCOR coefficient. In this case, differential decoding is also the same as this representation format.

(Second Embodiment)
FIG. 6 is a diagram illustrating the configuration of the speech encoding device 12 according to the second embodiment. The speech encoding device 12 is physically provided with a CPU, ROM, RAM, communication device, etc. (not shown), and this CPU is a predetermined computer program (for example, stored in the internal memory of the speech encoding device 12 such as a ROM). The computer program for performing the processing shown in the flowchart of FIG. 7 is loaded into the RAM and executed to control the speech encoding apparatus 12 in an integrated manner. The communication device of the audio encoding device 12 receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside.

  The speech encoding device 12 is functionally replaced by a linear prediction coefficient thinning unit 1j (prediction coefficient thinning means), linear prediction, instead of the filter strength parameter calculation unit 1f and the bitstream multiplexing unit 1g of the speech encoding device 11. A coefficient quantization unit 1k (prediction coefficient quantization unit) and a bit stream multiplexing unit 1g2 (bit stream multiplexing unit) are provided. The frequency conversion unit 1a to the linear prediction analysis unit 1e (linear prediction analysis unit), the linear prediction coefficient thinning unit 1j, the linear prediction coefficient quantization unit 1k, and the bit stream multiplexing unit 1g2 of the speech encoding device 12 illustrated in FIG. This is a function realized by the CPU of the speech encoding device 12 executing a computer program stored in the built-in memory of the speech encoding device 12. The CPU of the speech encoding device 12 executes this computer program (frequency conversion unit 1a to linear prediction analysis unit 1e, linear prediction coefficient thinning unit 1j, linear prediction coefficient quantum of the speech encoding device 12 shown in FIG. 6). 7 (using the conversion unit 1k and the bitstream multiplexing unit 1g2), the processes shown in the flowchart of FIG. 7 (steps Sa1 to Sa5 and steps Sc1 to Sc3) are sequentially executed. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech encoding device 12.

The linear prediction coefficient decimation unit 1j decimates a _H (n, r) obtained from the linear prediction analysis unit 1e in the time direction, and a value for a part of time slots r _i in a _H (n, r), The corresponding value of r _i is transmitted to the linear prediction coefficient quantization unit 1k (processing of step Sc1). However, 0 ≦ i <N _ts , where N _ts is the number of time slots in which a _H (n, r) is transmitted in the frame. The thinning out of the linear prediction coefficient may be based on a certain time interval, or may be thinned out based on the property of a _H (n, r). For example, _a H (n, r) in a frame having a length comparing _G H (r) _of, when _{the G} H (r) exceeds a predetermined value _a H (n, r) and A method such as a method for quantization is conceivable. The decimation interval of the linear prediction coefficients a _{H (n,} r) in the case of a constant distance regardless of the nature of, for that do not qualify time slot of transmission necessary to calculate a _{H (n,} r) There is no.

The linear prediction coefficient quantization unit 1k quantizes the thinned high-frequency linear prediction coefficient a _H (n, r _i ) given from the linear prediction coefficient thinning unit 1j and the index r _i of the corresponding time slot, and generates a bit stream. The data is transmitted to the multiplexing unit 1g2 (step Sc2 processing). As an alternative configuration, instead of quantizing a _H (n, r _i ), the linear prediction coefficient difference value a _D , as in the speech coding apparatus according to the second modification of the first embodiment. (N, r _i ) may be the target of quantization.

The bitstream multiplexing unit 1g2 includes the encoded bitstream calculated by the core codec encoding unit 1c, the SBR auxiliary information calculated by the SBR encoding unit 1d, and the quantum given from the linear prediction coefficient quantization unit 1k. The time slot index {r _i } corresponding to the converted a _H (n, r _i ) is multiplexed into a bit stream, and this multiplexed bit stream is output via the communication device of the speech encoding device 12. (Process of step Sc3).

  FIG. 8 is a diagram illustrating a configuration of the speech decoding apparatus 22 according to the second embodiment. The voice decoding device 22 includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated. The CPU is a predetermined computer program (for example, a diagram) stored in a built-in memory of the voice decoding device 22 such as a ROM. The speech decoding apparatus 22 is centrally controlled by loading a computer program for performing the processing shown in the flowchart of FIG. The communication device of the audio decoding device 22 receives the encoded multiplexed bit stream output from the audio encoding device 12, and further outputs the decoded audio signal to the outside.

  The speech decoding device 22 is functionally replaced by the bit stream separation unit 2a, the low frequency linear prediction analysis unit 2d, the signal change detection unit 2e, the filter strength adjustment unit 2f, and the linear prediction filter unit 2k of the speech decoding device 21. , A bit stream separation unit 2a1 (bit stream separation unit), a linear prediction coefficient interpolation / extrapolation 2p (linear prediction coefficient interpolation / extrapolation unit), and a linear prediction filter unit 2k1 (time envelope transformation unit). The bit stream separation unit 2a1, the core codec decoding unit 2b, the frequency conversion unit 2c, the high frequency generation unit 2g to the high frequency adjustment unit 2j, the linear prediction filter unit 2k1, the coefficient addition unit 2m, and the frequency inverse conversion of the speech decoding device 22 illustrated in FIG. The unit 2n and the linear prediction coefficient interpolation / external external 2p are functions realized by the CPU of the speech decoding device 22 executing a computer program stored in the internal memory of the speech decoding device 22. The CPU of the speech decoding device 22 executes this computer program (the bit stream separation unit 2a1, the core codec decoding unit 2b, the frequency conversion unit 2c, the high frequency generation unit 2g to the high frequency adjustment unit 2j, and linear prediction shown in FIG. 8). Filter unit 2k1, coefficient adding unit 2m, frequency inverse transform unit 2n, and linear prediction coefficient interpolation / complementary external 2p), processing shown in the flowchart of FIG. 9 (steps Sb1 to Sb2, step Sd1, and step Sb5) Steps Sb8, Sd2, and steps Sb10 to Sb11) are sequentially executed. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech decoding device 22.

  The speech decoding device 22 replaces the bit stream separation unit 2a, the low frequency linear prediction analysis unit 2d, the signal change detection unit 2e, the filter strength adjustment unit 2f, and the linear prediction filter unit 2k of the speech decoding device 22 with a bit stream separation unit. 2a1, linear prediction coefficient interpolation / external 2p, and linear prediction filter unit 2k1.

The bit stream separation unit 2a1 is configured to quantize the multiplexed bit stream input via the communication device of the audio decoding device 22 with the index r _{i of the} time slot corresponding to the quantized a _H (n, r _i ) and the SBR. The auxiliary information and the encoded bit stream are separated.

ただし、ｒ_ｉ０は線形予測係数が伝送されている時間スロット｛ｒ_ｉ｝のうちｒに最も近いものとする。また、δは０＜δ＜１を満たす定数である。 The linear prediction coefficient interpolation / extrapolation 2p receives the index r _i of the time slot corresponding to the quantized a _H (n, r _i ) from the bitstream separation unit 2a1, and receives the time slot in which no linear prediction coefficient is transmitted. A _H (n, r) corresponding to is acquired by interpolation or extrapolation (processing of step Sd1). The linear prediction coefficient interpolation / extrapolation 2p can perform extrapolation of the linear prediction coefficient, for example, according to the following equation (16).

However, r _i0 is the closest to r of the time the linear prediction coefficients are transmitted slots {r _i}. Also, δ is a constant that satisfies 0 <δ <1.

Further, the linear prediction coefficient interpolation / complementary external 2p can perform interpolation of the linear prediction coefficient in accordance with, for example, the following equation (17). However, r _i0 <r <r _{i0 + 1} is satisfied.

In addition, the linear prediction coefficient interpolation / external 2p is another expression format such as LSP (Linear Spectrum Pair), ISP (Immittance Spectrum Pair), LSF (Linear Spectrum Frequency), ISF (Immittance Spectrum Frequency), and PARCOR coefficient. After conversion to, interpolation and extrapolation may be performed, and the obtained value may be converted into a linear prediction coefficient. The interpolated or extrapolated a _H (n, r) is transmitted to the linear prediction filter unit 2k1 and used as a linear prediction coefficient in the linear prediction synthesis filter process, but used as a linear prediction coefficient in the linear prediction inverse filter unit 2i. May be. When a _D (n, r _i ) is multiplexed in the bitstream instead of a _H (n, r), the linear prediction coefficient interpolation / extrapolation 2p performs the first step prior to the above interpolation or extrapolation processing. The same differential decoding process as that of the speech decoding apparatus according to the second modification of the embodiment is performed.

The linear prediction filter unit 2k1 interpolates or extrapolates a _H (n, r) obtained from the linear prediction coefficient interpolation / extrapolation 2p with respect to q _adj (n, r) output from the high frequency adjustment unit 2j. ) Is used to perform linear prediction synthesis filter processing in the frequency direction (step Sd2 processing). The transfer function of the linear prediction filter unit 2k1 is as the following formula (18). Similar to the linear prediction filter unit 2k of the speech decoding apparatus 21, the linear prediction filter unit 2k1 performs a linear prediction synthesis filter process to transform the time envelope of the high-frequency component generated by SBR.

(Third embodiment)
FIG. 10 is a diagram illustrating a configuration of the speech encoding device 13 according to the third embodiment. The speech encoding device 13 is physically provided with a CPU, ROM, RAM, communication device, and the like (not shown). The computer program for performing the processing shown in the flowchart of FIG. 11 is loaded into the RAM and executed to control the speech encoding apparatus 13 in an integrated manner. The communication device of the audio encoding device 13 receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside.

  The speech encoding device 13 functionally replaces the linear prediction analysis unit 1e, the filter strength parameter calculation unit 1f, and the bit stream multiplexing unit 1g of the speech encoding device 11 in terms of a time envelope calculation unit 1m (time envelope assist). Information calculation unit), an envelope shape parameter calculation unit 1n (temporal envelope auxiliary information calculation unit), and a bit stream multiplexing unit 1g3 (bit stream multiplexing unit). The frequency converters 1a to SBR encoder 1d, the time envelope calculator 1m, the envelope shape parameter calculator 1n, and the bit stream multiplexer 1g3 of the speech encoder 13 shown in FIG. This is a function realized by the CPU executing a computer program stored in the built-in memory of the speech encoding device 13. The CPU of the speech coder 13 executes this computer program (frequency converters 1a to SBR coder 1d, time envelope calculator 1m, envelope shape parameter calculator of the speech coder 13 shown in FIG. 10). 1n and the bit stream multiplexing unit 1g3), the processes shown in the flowchart of FIG. 11 (steps Sa1 to Sa4 and steps Se1 to Se3) are sequentially executed. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech encoding device 13.

The time envelope calculation unit 1m receives q (k, r) and acquires time envelope information e (r) of a high frequency component of the signal by acquiring power for each time slot of q (k, r), for example. (Step Se1 processing). In this case, e (r) is obtained according to the following mathematical formula (19).

ただし、

上記の数式におけるｓ（ｉ）はｂ_ｉ≦ｒ＜ｂ_ｉ＋１を満たすｉ番目のＳＢＲエンベロープ内におけるｅ（ｒ）の変化の大きさを示すパラメータであり、時間エンベロープの変化が大きいほどｅ（ｒ）は大きい値をとる。上記の数式（２０）及び（２１）は、ｓ（ｉ）の算出方法の一例であり、例えばｅ（ｒ）のＳＭＦ（Spectral Flatness Measure）や、最大値と最小値の比等、を用いてｓ（ｉ）を取得してもよい。この後、ｓ（ｉ）は量子化され、ビットストリーム多重化部１ｇ３に伝送される。 The envelope shape parameter calculation unit 1n receives e (r) from the time envelope calculation unit 1m, and further receives a time boundary {b _i } of the SBR envelope from the SBR encoding unit 1d. However, 0 ≦ i ≦ Ne, and Ne is the number of SBR envelopes in the encoded frame. The envelope shape parameter calculation unit 1n acquires the envelope shape parameter s (i) (0 ≦ i <Ne), for example, according to the following equation (20) for each of the SBR envelopes in the encoded frame (processing of step Se2). . Note that the envelope shape parameter s (i) corresponds to the time envelope auxiliary information and is the same in the third embodiment.

However,

In the above formula, s (i) is a parameter indicating the magnitude of change of e (r) in the i-th SBR envelope that satisfies b _i ≦ r <b _{i + 1} , and e (r ) Takes a large value. The above mathematical formulas (20) and (21) are examples of the calculation method of s (i), and for example, using SMF (Spectral Flatness Measure) of e (r), the ratio between the maximum value and the minimum value, and the like. s (i) may be acquired. Thereafter, s (i) is quantized and transmitted to the bitstream multiplexing unit 1g3.

  The bitstream multiplexing unit 1g3 multiplexes the encoded bitstream calculated by the core codec encoding unit 1c, the SBR auxiliary information calculated by the SBR encoding unit 1d, and s (i) into the bitstream, The multiplexed bit stream is output via the communication device of the speech encoding device 13 (processing of step Se3).

  FIG. 12 is a diagram illustrating a configuration of the speech decoding apparatus 23 according to the third embodiment. The speech decoding device 23 includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a predetermined computer program (for example, FIG. The computer program for performing the processing shown in the flowchart in FIG. The communication device of the audio decoding device 23 receives the encoded multiplexed bit stream output from the audio encoding device 13, and further outputs the decoded audio signal to the outside.

  The speech decoding device 23 functionally includes a bit stream separation unit 2a, a low frequency linear prediction analysis unit 2d, a signal change detection unit 2e, a filter strength adjustment unit 2f, a high frequency linear prediction analysis unit 2h, and a linear function. Instead of the prediction inverse filter unit 2i and the linear prediction filter unit 2k, a bit stream separation unit 2a2 (bit stream separation unit), a low frequency time envelope calculation unit 2r (low frequency time envelope analysis unit), and an envelope shape adjustment unit 2s (time Envelope adjusting means), a high-frequency time envelope calculating section 2t, a time envelope flattening section 2u, and a time envelope deforming section 2v (time envelope deforming means). The bit stream separation unit 2a2, the core codec decoding unit 2b to the frequency conversion unit 2c, the high frequency generation unit 2g, the high frequency adjustment unit 2j, the coefficient addition unit 2m, the frequency inverse conversion unit 2n, and the low The frequency time envelope calculation unit 2r to the time envelope transformation unit 2v are functions realized by the CPU of the speech decoding device 23 executing a computer program stored in the internal memory of the speech decoding device 23. The CPU of the speech decoding device 23 executes this computer program (the bit stream separation unit 2a2, the core codec decoding unit 2b to the frequency conversion unit 2c, the high frequency generation unit 2g, and the high frequency adjustment of the speech decoding device 23 shown in FIG. 12). Unit 2j, coefficient addition unit 2m, frequency inverse conversion unit 2n, and low frequency time envelope calculation unit 2r to time envelope transformation unit 2v), and the processing shown in the flowchart of FIG. 13 (steps Sb1 to Sb2, step Sf1) Step Sf2, Step Sb5, Step Sf3 to Step Sf4, Step Sb8, Step Sf5, and Step Sb10 to Step Sb11) are sequentially executed. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech decoding device 23.

The bit stream separation unit 2a2 separates the multiplexed bit stream input via the communication device of the audio decoding device 23 into s (i), SBR auxiliary information, and an encoded bit stream. The low frequency time envelope calculation unit 2r receives q _dec (k, r) including the low frequency component from the frequency conversion unit 2c, and acquires e (r) according to the following equation (22) (processing of step Sf1).

ただし、

である。 The envelope shape adjusting unit 2s adjusts e (r) using s (i), and acquires adjusted time envelope information e _adj (r) (processing in step Sf2). The adjustment to e (r) can be performed, for example, according to the following mathematical formulas (23) to (25).

However,

It is.

The above formulas (23) to (25) are examples of adjustment methods, and other adjustment methods may be used such that the shape of e _adj (r) approaches the shape indicated by s (i).

The high frequency time envelope calculation unit 2t calculates the time envelope e _exp (r) according to the following equation (26) using q _exp (k, r) obtained from the high frequency generation unit 2g (processing in step Sf3).

The time envelope flattening unit 2u flattens the time envelope of q _exp (k, r) obtained from the high frequency generation unit 2g according to the following equation (27), and the obtained signal q _flat (k, r) in the QMF region. ) Is transmitted to the high frequency adjustment unit 2j (processing of step Sf4).

The flattening of the time envelope in the time envelope flattening unit 2u may be omitted. Further, instead of performing the time envelope calculation of the high frequency component and the flattening process of the time envelope on the output from the high frequency generation unit 2g, the time envelope calculation of the high frequency component is performed on the output from the high frequency adjustment unit 2j. Time envelope flattening processing may be performed. Furthermore, the time envelope used in the time envelope flattening unit 2u is not e _exp (r) obtained from the high frequency time envelope calculating unit 2t, but e _edj (r) obtained from the envelope shape adjusting unit 2s. Good.

The time envelope deformation unit 2v deforms q _adj (k, r) obtained from the high frequency adjustment unit 2j using e _adj (r) obtained from the time envelope deformation unit 2v, and the QMF whose time envelope is deformed. The signal _qenvadj (k, r) of the area is acquired (processing of step Sf5). This deformation is performed according to the following equation (28). _qenvadj (k, r) is transmitted to the coefficient adding unit 2m as a signal in the QMF region corresponding to the high frequency component.

(Fourth embodiment)
FIG. 14 is a diagram showing the configuration of the speech decoding apparatus 24 according to the fourth embodiment. The voice decoding device 24 is physically provided with a CPU, ROM, RAM, a communication device, etc. (not shown), and this CPU loads a predetermined computer program stored in the internal memory of the voice decoding device 24 such as a ROM into the RAM. The speech decoding device 24 is controlled in an integrated manner. The communication device of the audio decoding device 24 receives the encoded multiplexed bit stream output from the audio encoding device 11 or the audio encoding device 13, and further outputs the decoded audio signal to the outside.

  The speech decoding device 24 functionally includes the configuration of the speech decoding device 21 (core codec decoding unit 2b, frequency conversion unit 2c, low frequency linear prediction analysis unit 2d, signal change detection unit 2e, filter strength adjustment unit 2f, high frequency The configuration (low) of the generation unit 2g, the high-frequency linear prediction analysis unit 2h, the linear prediction inverse filter unit 2i, the high-frequency adjustment unit 2j, the linear prediction filter unit 2k, the coefficient addition unit 2m, and the frequency inverse conversion unit 2n) A frequency time envelope calculation unit 2r, an envelope shape adjustment unit 2s, and a time envelope deformation unit 2v). Furthermore, the audio decoding device 24 includes a bit stream separation unit 2a3 (bit stream separation unit) and an auxiliary information conversion unit 2w. The order of the linear prediction filter unit 2k and the time envelope transformation unit 2v may be the reverse of that shown in FIG. Note that the speech decoding device 24 preferably receives a bit stream encoded by the speech encoding device 11 or the speech encoding device 13 as an input. The configuration of the speech decoding device 24 shown in FIG. 14 is a function realized by the CPU of the speech decoding device 24 executing a computer program stored in the built-in memory of the speech decoding device 24. It is assumed that various data necessary for the execution of the computer program and various data generated by the execution of the computer program are all stored in a built-in memory such as a ROM or a RAM of the speech decoding device 24.

  The bit stream separation unit 2a3 separates the multiplexed bit stream input via the communication device of the audio decoding device 24 into time envelope auxiliary information, SBR auxiliary information, and an encoded bit stream. The time envelope auxiliary information may be K (r) described in the first embodiment or s (i) described in the third embodiment. Further, it may be another parameter X (r) that is neither K (r) nor s (i).

を取得した後に、所定のテーブルを用いて、この数式（２９）に示す平均値をｓ（ｉ）に変換することによって行ってもよい。また、時間エンベロープ補助情報がｓ（ｉ）の場合、補助情報変換部２ｗは、ｓ（ｉ）をＫ（ｒ）に変換する。補助情報変換部２ｗは、この変換を、例えば所定のテーブルを用いてｓ（ｉ）をＫ（ｒ）に変換することによって行ってもよい。ただし、ｉとｒはｂ_ｉ≦ｒ＜ｂ_ｉ＋１の関係を満たすよう対応づけられるものとする。 The auxiliary information conversion unit 2w converts the input time envelope auxiliary information to obtain K (r) and s (i). When the time envelope auxiliary information is K (r), the auxiliary information conversion unit 2w converts K (r) to s (i). The auxiliary information conversion unit 2w performs this conversion, for example, an average value of K (r) in a section of b _i ≦ r <b _{i + 1.}

May be obtained by converting the average value shown in Equation (29) into s (i) using a predetermined table. When the time envelope auxiliary information is s (i), the auxiliary information conversion unit 2w converts s (i) to K (r). The auxiliary information conversion unit 2w may perform this conversion by converting s (i) to K (r) using a predetermined table, for example. However, i and r shall be matched so as to satisfy the relationship of b _i ≦ r <b _{i + 1} .

  When the time envelope auxiliary information is a parameter X (r) that is neither s (i) nor K (r), the auxiliary information conversion unit 2w converts X (r) into K (r) and s (i). . The auxiliary information conversion unit 2w desirably performs this conversion by converting X (r) into K (r) and s (i) using a predetermined table, for example. The auxiliary information conversion unit 2w preferably transmits one representative value for each SBR envelope. The tables for converting X (r) into K (r) and s (i) may be different from each other.

ここで、Ｐ_０（ｒ），Ｐ_１（ｒ）はそれぞれ以下の数式（３１）及び数式（３２）で表される。

この自動利得制御処理により、線形予測フィルタ部２ｋの出力信号の高周波成分の電力は線形予測フィルタ処理前と等しい値に調整される。その結果、ＳＢＲに基づいて生成された高周波成分の時間エンベロープを変形した線形予測フィルタ部２ｋの出力信号において、高周波調整部２ｊにおいて行われた高周波信号の電力の調整の効果が保たれる。なお、この自動利得制御処理は，ＱＭＦ領域の信号の任意の周波数範囲に対して個別に行うことも可能である。個々の周波数範囲に対する処理は、それぞれ、数式（３０）、数式（３１）、数式（３２）のｎをある周波数範囲に限定することで実現できる。例えばｉ番目の周波数範囲はＦ_ｉ≦ｎ＜Ｆ_ｉ＋１と表すことができる（この場合のｉは、ＱＭＦ領域の信号の任意の周波数範囲の番号を示すインデックスである）。Ｆ_ｉは周波数範囲の境界を示し、“MPEG4 AAC”のＳＢＲにおいて規定されるエンベロープスケールファクタの周波数境界テーブルであることが望ましい。周波数境界テーブルは“MPEG4 AAC”のＳＢＲの規定に従い、高周波生成部２ｇにおいて決定される。この自動利得制御処理により、線形予測フィルタ部２ｋの出力信号の高周波成分の任意の周波数範囲内の電力は線形予測フィルタ処理前と等しい値に調整される。その結果、ＳＢＲに基づいて生成された高周波成分の時間エンベロープを変形した線形予測フィルタ部２ｋの出力信号で、高周波調整部２ｊにおいて行われた高周波信号の電力の調整の効果が周波数範囲の単位で保たれる。また、第１の実施形態の本変形例３と同様の変更を第４の実施形態における線形予測フィルタ部２ｋに加えてもよい。 (Modification 3 of the first embodiment)
In the speech decoding device 21 of the first embodiment, the linear prediction filter unit 2k of the speech decoding device 21 can include an automatic gain control process. This automatic gain control process is a process for matching the power of the QMF domain signal output from the linear prediction filter unit 2k to the input signal power of the QMF domain. The QMF region signal q _{syn, pow} (n, r) after gain control is generally realized by the following equation.

Here, P ₀ (r) and P ₁ (r) are represented by the following formulas (31) and (32), respectively.

By this automatic gain control processing, the power of the high frequency component of the output signal of the linear prediction filter unit 2k is adjusted to a value equal to that before the linear prediction filter processing. As a result, the effect of adjusting the power of the high-frequency signal performed in the high-frequency adjusting unit 2j is maintained in the output signal of the linear prediction filter unit 2k obtained by modifying the time envelope of the high-frequency component generated based on the SBR. Note that this automatic gain control processing can be performed individually for an arbitrary frequency range of a signal in the QMF region. The processing for each frequency range can be realized by limiting n in Equation (30), Equation (31), and Equation (32) to a certain frequency range, respectively. For example, the i-th frequency range can be expressed as F _i ≦ n <F _{i + 1} (where i is an index indicating the number of an arbitrary frequency range of the signal in the QMF domain). F _i represents a frequency range boundary, and is preferably a frequency boundary table of an envelope scale factor defined in the SBR of “MPEG4 AAC”. The frequency boundary table is determined by the high frequency generator 2g in accordance with the SBR specification of “MPEG4 AAC”. By this automatic gain control process, the power within an arbitrary frequency range of the high-frequency component of the output signal of the linear prediction filter unit 2k is adjusted to a value equal to that before the linear prediction filter process. As a result, the effect of adjusting the power of the high-frequency signal performed in the high-frequency adjusting unit 2j in the output signal of the linear prediction filter unit 2k obtained by modifying the time envelope of the high-frequency component generated based on the SBR is in units of frequency range. Kept. Moreover, you may add the change similar to this modification 3 of 1st Embodiment to the linear prediction filter part 2k in 4th Embodiment.

ただし、

はe（ｒ）のＳＢＲエンベロープ内での平均値であり、その算出方法は数式（２１）に従う。ただし、ＳＢＲエンベロープとは、ｂ_ｉ≦ｒ＜ｂ_ｉ＋１を満たす時間範囲を示す。また、｛ｂ_ｉ｝は、ＳＢＲ補助情報に情報として含まれている、ＳＢＲエンベロープの時間境界であり、任意の時間範囲、任意の周波数範囲の平均信号エネルギーを表すＳＢＲエンベロープスケールファクタが対象とする時間範囲の境界である。また、ｍｉｎ（・）はｂ_ｉ≦ｒ＜ｂ_ｉ＋１の範囲における最小値を表す。従って、この場合には、エンベロープ形状パラメータｓ（ｉ）は、調整後の時間エンベロープ情報のＳＢＲエンベロープ内での最小値と平均値の比率を指示するパラメータである。また、第３の実施形態の音声復号装置２３におけるエンベロープ形状調整部２ｓは、以下のような処理で実現することもできる。エンベロープ形状調整部２ｓは、ｓ（ｉ）を用いてｅ（ｒ）を調整し、調整後の時間エンベロープ情報ｅ_ａｄｊ（ｒ）を取得する。調整の方法は次の数式（３５）又は数式（３６）に従う。

数式３５は、調整後の時間エンベロープ情報ｅ_ａｄｊ（ｒ）のＳＢＲエンベロープ内での最小値と平均値の比率が、エンベロープ形状パラメータｓ（ｉ）の値と等しくなるようエンベロープ形状を調整するものである。また、上記した第３の実施形態の本変形例１と同様の変更を第４の実施形態に加えてもよい。 (Modification 1 of 3rd Embodiment)
The envelope shape parameter calculation unit 1n in the speech encoding device 13 according to the third embodiment can also be realized by the following processing. The envelope shape parameter calculation unit 1n obtains the envelope shape parameter s (i) (0 ≦ i <Ne) for each of the SBR envelopes in the encoded frame according to the following equation (33).

However,

Is the average value of e (r) within the SBR envelope, and the calculation method follows Formula (21). However, the SBR envelope indicates a time range that satisfies b _i ≦ r <b _{i + 1} . {B _i } is a time boundary of the SBR envelope included as information in the SBR auxiliary information, and is targeted for an SBR envelope scale factor representing the average signal energy in an arbitrary time range and an arbitrary frequency range. It is the boundary of the time range. Min (·) represents the minimum value in the range of b _i ≦ r <b _{i + 1} . Therefore, in this case, the envelope shape parameter s (i) is a parameter that indicates the ratio between the minimum value and the average value in the SBR envelope of the adjusted time envelope information. Further, the envelope shape adjusting unit 2s in the speech decoding apparatus 23 according to the third embodiment can also be realized by the following processing. The envelope shape adjusting unit 2s adjusts e (r) using s (i), and obtains time envelope information e _adj (r) after adjustment. The adjustment method follows the following formula (35) or formula (36).

Equation 35 adjusts the envelope shape so that the ratio between the minimum value and the average value in the SBR envelope of the adjusted time envelope information e _adj (r) is equal to the value of the envelope shape parameter s (i). is there. Moreover, you may add the same change as this modification 1 of 3rd Embodiment mentioned above to 4th Embodiment.

また、上記した第３の実施形態の本変形例２と同様の変更を第４の実施形態に加えてもよい。 (Modification 2 of the third embodiment)
The time envelope deforming unit 2v can use the following formula instead of the formula (28). As shown in Equation (37), e _{adj, scaled} (r) is the time envelope information e after adjustment so that the power in the SBR envelope of q _adj (k, r) and q _envadj (k, r) are equal. The gain of _adj (r) is controlled. Further, as shown in equation (38), in the second modification of the third _{embodiment, e adj} (r) rather than _{e adj,} multiply _scaled to (r) signal _q adj (k, r) of the QMF region To obtain q _envadj (k, r). Therefore, the time envelope deforming unit 2v can deform the time envelope of the signal _qadj (k, r) in the QMF region so that the signal power in the SBR envelope becomes equal before and after the deformation of the time envelope. it can. However, the SBR envelope indicates a time range that satisfies b _i ≦ r <b _{i + 1} . {B _i } is a time boundary of the SBR envelope included as information in the SBR auxiliary information, and is targeted for an SBR envelope scale factor representing the average signal energy in an arbitrary time range and an arbitrary frequency range. It is the boundary of the time range. The term “SBR envelope” in the embodiments of the present invention corresponds to the term “SBR envelope time segment” in “MPEG4 AAC” defined in “ISO / IEC 14496-3”. “Envelope” means the same content as “SBR envelope time segment”.

Moreover, you may add the change similar to this modification 2 of 3rd Embodiment mentioned above to 4th Embodiment.

数式（２２）は下記の数式（４０）であってもよい。

数式（２６）は下記の数式（４１）であってもよい。

数式（３９）及び数式（４０）にしたがった場合、時間エンベロープ情報ｅ（ｒ）は、ＱＭＦサブバンドサンプルごとの電力をＳＢＲエンベロープ内での平均電力で正規化し、さらに平方根をとったものとなる。ただし、ＱＭＦサブバンドサンプルは、ＱＭＦ領域信号において、同一の時間インデックス“ｒ”に対応する信号ベクトルであり、QMF領域における一つのサブサンプルを意味する。また、本発明の実施形態全体において、用語”時間スロット”は”ＱＭＦサブバンドサンプル”と同一の内容を意味する。この場合、時間エンベロープ情報ｅ（ｒ）は、各ＱＭＦサブバンドサンプルへ乗算されるべきゲイン係数を意味することとなり、調整後の時間エンベロープ情報ｅ_ａｄｊ（ｒ）も同様である。 (Modification 3 of the third embodiment)
The mathematical formula (19) may be the following mathematical formula (39).

The mathematical formula (22) may be the following mathematical formula (40).

The mathematical formula (26) may be the following mathematical formula (41).

In accordance with Equation (39) and Equation (40), the time envelope information e (r) is obtained by normalizing the power for each QMF subband sample with the average power in the SBR envelope and taking the square root. . However, the QMF subband sample is a signal vector corresponding to the same time index “r” in the QMF domain signal, and means one subsample in the QMF domain. Also, throughout the embodiments of the present invention, the term “time slot” means the same content as “QMF subband sample”. In this case, the time envelope information e (r) means a gain coefficient to be multiplied to each QMF subband sample, and the adjusted time envelope information e _adj (r) is the same.

(Modification 1 of 4th Embodiment)
A speech decoding device 24a (not shown) of Modification 1 of the fourth embodiment includes a CPU, ROM, RAM, communication device, and the like which are not physically shown. A predetermined computer program stored in the built-in memory is loaded into the RAM and executed, thereby comprehensively controlling the speech decoding device 24a. The communication device of the audio decoding device 24a receives the encoded multiplexed bit stream output from the audio encoding device 11 or the audio encoding device 13, and further outputs the decoded audio signal to the outside. The audio decoding device 24a functionally includes a bit stream separation unit 2a4 (not shown) instead of the bit stream separation unit 2a3 of the audio decoding device 24, and further replaces the auxiliary information conversion unit 2w with time envelope auxiliary information. A generation unit 2y (not shown) is provided. The bit stream separation unit 2a4 separates the multiplexed bit stream into SBR auxiliary information and an encoded bit stream. The time envelope auxiliary information generation unit 2y generates time envelope auxiliary information based on information included in the encoded bitstream and the SBR auxiliary information.

For generating the time envelope auxiliary information in a certain SBR envelope, for example, the time width (b _{i + 1} −b _i ) of the SBR envelope, the frame class, the strength parameter of the inverse filter, the noise floor, the magnitude of the high frequency power, the high frequency power and the low The ratio of frequency power, the autocorrelation coefficient or the prediction gain as a result of linear prediction analysis of the low frequency signal expressed in the QMF region in the frequency direction can be used. The time envelope auxiliary information can be generated by determining K (r) or s (i) based on one or more values of these parameters. For example the time width of SBR envelopes _{(b i} + 1 -b _i) larger the K (r) or s (i) such decrease, or SBR envelope time width _{(b i} + 1 -b _i) larger the K (r) Alternatively, time envelope auxiliary information can be generated by determining K (r) or s (i) based on (b _{i + 1} −b _i ) so that s (i) becomes large. Moreover, you may add the same change to 1st Embodiment and 3rd Embodiment.

(Modification 2 of the fourth embodiment)
The speech decoding device 24b (see FIG. 15) of Modification 2 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech decoding device 24b such as a ROM. A predetermined computer program stored in the built-in memory is loaded into the RAM and executed to control the speech decoding apparatus 24b in an integrated manner. The communication device of the audio decoding device 24b receives the encoded multiplexed bit stream output from the audio encoding device 11 or the audio encoding device 13, and further outputs the decoded audio signal to the outside. As shown in FIG. 15, the speech decoding device 24b includes a primary high frequency adjustment unit 2j1 and a secondary high frequency adjustment unit 2j2 instead of the high frequency adjustment unit 2j.

Here, the primary high frequency adjustment unit 2j1 performs linear prediction inverse filter processing in a time direction, gain adjustment, and noise superimposition processing for a signal in the QMF region of the high frequency band in the “HF adjustment” step in the SBR of “MPEG4 AAC” Make adjustments with. At this time, the output signal of the primary high frequency adjusting unit 2j1 is, "ISO / IEC14496-3: 2005" of "SBR tool" in, equivalent to the signal W ₂ in the description of 4.6.18.7.6 Section "Assembling HFsignals" It becomes. The linear prediction filter unit 2k (or the linear prediction filter unit 2k1) and the time envelope deformation unit 2v perform time envelope deformation on the output signal of the primary high frequency adjustment unit. The secondary high frequency adjustment unit 2j2 performs a sine wave addition process in the “HF adjustment” step in the SBR of “MPEG4 AAC” on the QMF domain signal output from the time envelope transformation unit 2v. Treatment of the secondary high frequency adjusting section, "ISO / IEC14496-3: 2005" in the "SBR tool", within the description of 4.6.18.7.6 Section "Assembling HFsignals", the process of generating a signal Y from the signal W _{2 2} corresponds to the processing in which the signal W2 is replaced with the output signal of the time envelope deforming unit 2v.

  In the above description, only the sine wave addition process is the process of the secondary high frequency adjustment unit 2j2, but any of the processes in the “HF adjustment” step may be the process of the secondary high frequency adjustment unit 2j2. Moreover, you may add the same deformation | transformation to 1st Embodiment, 2nd Embodiment, and 3rd Embodiment. At this time, since the first embodiment and the second embodiment include the linear prediction filter units (linear prediction filter units 2k and 2k1) and do not include the time envelope deformation unit, the output signal of the primary high frequency adjustment unit 2j1 After the processing in the linear prediction filter unit, the processing in the secondary high frequency adjustment unit 2j2 is performed on the output signal of the linear prediction filter unit.

  In addition, since the third embodiment includes the time envelope deforming unit 2v and does not include the linear prediction filter unit, the time envelope deforming unit 2v performs processing on the output signal of the primary high frequency adjusting unit 2j1, and then the time The secondary high frequency adjustment unit performs processing on the output signal of the envelope deformation unit 2v.

  Further, in the speech decoding device (speech decoding devices 24, 24a, and 24b) of the fourth embodiment, the order of processing of the linear prediction filter unit 2k and the time envelope transformation unit 2v may be reversed. That is, the processing of the time envelope deforming unit 2v is first performed on the output signal of the high frequency adjusting unit 2j or the primary high frequency adjusting unit 2j1, and then the linear prediction filter unit 2k is output on the output signal of the time envelope deforming unit 2v. You may perform the process of.

  The temporal envelope auxiliary information includes binary control information for instructing whether or not to perform processing in the linear prediction filter unit 2k or the temporal envelope transformation unit 2v, and this control information is the linear prediction filter unit 2k or temporal envelope transformation. Only when it is instructed to perform processing in the section 2v, a filter strength parameter K (r), an envelope shape parameter s (i), or a parameter that determines both K (r) and s (i) It may take a form that further includes any one or more of X (r) as information.

(Modification 3 of the fourth embodiment)
A speech decoding device 24c (see FIG. 16) of Modification 3 of the fourth embodiment includes a CPU, ROM, RAM, communication device, and the like which are not shown physically, and this CPU is a speech decoding device 24c such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 17) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24c in an integrated manner. The communication device of the audio decoding device 24c receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 16, the speech decoding device 24c includes a primary high frequency adjustment unit 2j3 and a secondary high frequency adjustment unit 2j4 in place of the high frequency adjustment unit 2j, and further replaces the linear prediction filter unit 2k and the time envelope modification unit 2v. Individual signal component adjustment units 2z1, 2z2, and 2z3 (the individual signal component adjustment unit corresponds to a time envelope deforming unit).

  The primary high frequency adjustment unit 2j3 outputs a signal in the QMF region in the high frequency band as a copy signal component. The primary high frequency adjustment unit 2j3 uses the SBR auxiliary information provided from the bitstream separation unit 2a3 for the signal in the QMF region in the high frequency band and performs linear prediction inverse filter processing in the time direction and gain adjustment (frequency characteristic adjustment). ) May be output as a copy signal component. Further, the primary high frequency adjustment unit 2j3 generates a noise signal component and a sine wave signal component using the SBR auxiliary information given from the bit stream separation unit 2a3, and separates the copy signal component, the noise signal component, and the sine wave signal component. Each of them is output in the form (process of step Sg1). The noise signal component and the sine wave signal component may depend on the content of the SBR auxiliary information and may not be generated.

  The individual signal component adjustment units 2z1, 2z2, and 2z3 perform processing on each of the plurality of signal components included in the output of the primary high frequency adjustment unit (processing in step Sg2). The processing in the individual signal component adjustment units 2z1, 2z2, 2z3 may be linear prediction synthesis filter processing in the frequency direction using the linear prediction coefficient obtained from the filter strength adjustment unit 2f, similar to the linear prediction filter unit 2k. Good (processing 1). Further, the processing in the individual signal component adjustment units 2z1, 2z2, and 2z3 is similar to the time envelope deformation unit 2v, and multiplies each QMF subband sample by a gain coefficient using the time envelope obtained from the envelope shape adjustment unit 2s. It may be a process (process 2). Further, the processing in the individual signal component adjustment units 2z1, 2z2, and 2z3 is linear prediction in the frequency direction using the linear prediction coefficient obtained from the filter strength adjustment unit 2f, similar to the linear prediction filter unit 2k, for the input signal. After performing the synthesis filter process, the QMF subband sample is multiplied by a gain coefficient using the time envelope obtained from the envelope shape adjusting unit 2s, similar to the time envelope deforming unit 2v, for the output signal. (Processing 3). The processing in the individual signal component adjustment units 2z1, 2z2, and 2z3 is performed on each QMF subband sample using the time envelope obtained from the envelope shape adjustment unit 2s similar to the time envelope deformation unit 2v for the input signal. After performing the process of multiplying the gain coefficient, the output signal is further subjected to linear prediction synthesis filter processing in the frequency direction using the linear prediction coefficient obtained from the filter strength adjustment unit 2f, similar to the linear prediction filter unit 2k. (Processing 4). The individual signal component adjustment units 2z1, 2z2, and 2z3 may output the input signal as it is without performing the time envelope transformation process on the input signal (processing 5). Also, the individual signal component adjustment unit 2z1 , 2z2, and 2z3 may add some processing for transforming the time envelope of the input signal by a method other than the processing 1 to 5 (processing 6). Further, the process in the individual signal component adjustment units 2z1, 2z2, and 2z3 may be a process in which a plurality of processes among the processes 1 to 6 are combined in an arbitrary order (process 7).

  The processing in the individual signal component adjustment units 2z1, 2z2, and 2z3 may be the same, but the individual signal component adjustment units 2z1, 2z2, and 2z3 are different from each other for each of the plurality of signal components included in the output of the primary high frequency adjustment unit. The time envelope may be modified by the method. For example, the individual signal component adjustment unit 2z1 performs processing 2 on the input copy signal, the individual signal component adjustment unit 2z2 performs processing 3 on the input noise signal component, and the individual signal component adjustment unit 2z3 is input. Different processes may be performed on each of the copy signal, the noise signal, and the sine wave signal, such as performing process 5 on the sine wave signal. At this time, the filter strength adjustment unit 2f and the envelope shape adjustment unit 2s may transmit the same linear prediction coefficient and time envelope to each of the individual signal component adjustment units 2z1, 2z2, and 2z3. Different linear prediction coefficients and time envelopes may be transmitted, and the same linear prediction coefficient and time envelope may be transmitted to any two or more of the individual signal component adjustment units 2z1, 2z2, and 2z3. One or more of the individual signal component adjustment units 2z1, 2z2, and 2z3 may output the input signal as it is without performing the time envelope transformation process (processing 5). Therefore, the individual signal component adjustment units 2z1, 2z2 , 2z3 as a whole performs time envelope processing on at least one of the plurality of signal components output from the primary high frequency adjustment unit 2j3 (all of the individual signal component adjustment units 2z1, 2z2, 2z3 are processing 5). In some cases, the time envelope deformation process is not performed for any signal component, and thus the present invention is not effective.

  The processing in each of the individual signal component adjustment units 2z1, 2z2, and 2z3 may be fixed to any one of the processing 1 to the processing 7, but any one of the processing 1 to the processing 7 is performed based on control information given from the outside. It may be determined dynamically whether or not to perform. At this time, the control information is preferably included in the multiplexed bit stream. Further, the control information may indicate whether to perform the processing 1 to the processing 7 in a specific SBR envelope time segment, an encoded frame, or other time range, and the control time range. The process 1 to the process 7 may be instructed without specifying.

  The secondary high-frequency adjusting unit 2j4 adds the processed signal components output from the individual signal component adjusting units 2z1, 2z2, and 2z3, and outputs the sum to the coefficient adding unit (processing in step Sg3). Further, the secondary high frequency adjustment unit 2j4 uses the SBR auxiliary information provided from the bit stream separation unit 2a3 for the copy signal component, and performs linear prediction inverse filter processing in the time direction and gain adjustment (frequency characteristic adjustment). You may perform at least one of these.

  The individual signal component adjustment units 2z1, 2z2, and 2z3 operate in cooperation with each other, add two or more signal components after performing any one of the processings 1 to 7 to each other, and Further, any one of the processes 1 to 7 may be added to generate an intermediate stage output signal. At this time, the secondary high-frequency adjusting unit 2j4 adds the intermediate stage output signal and the signal component not yet added to the intermediate stage output signal, and outputs the result to the coefficient adding unit. Specifically, the process 5 is performed on the copy signal component, the process 1 is added to the noise component, and then the two signal components are added to each other, and the process 2 is further added to the added signal. It is desirable to generate an output signal. At this time, the secondary high-frequency adjustment unit 2j4 adds the sine wave signal component to the output signal in the middle stage and outputs it to the coefficient addition unit.

  The primary high frequency adjustment unit 2j3 is not limited to the three signal components of the copy signal component, the noise signal component, and the sine wave signal component, and may output a plurality of arbitrary signal components in a separated form. In this case, the signal component may be a combination of two or more of a copy signal component, a noise signal component, and a sine wave signal component. Further, it may be a signal obtained by dividing one of a copy signal component, a noise signal component, and a sine wave signal component. The number of signal components may be other than 3, and in this case, the number of individual signal component adjustment units may be other than 3.

  The high-frequency signal generated by SBR is composed of three elements: a copy signal component obtained by copying a low-frequency band to a high-frequency band, a noise signal, and a sine wave signal. Since each of the copy signal, the noise signal, and the sine wave signal has a different time envelope, the time envelope is deformed in a different manner for each signal component as the individual signal component adjustment unit of the present modification performs. By performing the above, the subjective quality of the decoded signal can be further improved as compared with the other embodiments of the present invention. In particular, a noise signal generally has a flat time envelope, and a copy signal has a time envelope close to that of a low-frequency band signal. The time envelope can be controlled independently, which is effective in improving the subjective quality of the decoded signal. Specifically, a process (process 3 or process 4) for deforming the time envelope is performed on the noise signal, and a process (process 1 or process 2) different from that for the noise signal is performed on the copy signal. In addition, it is preferable to perform the process 5 on the sine wave signal (that is, do not perform the time envelope deformation process). Alternatively, time envelope deformation processing (processing 3 or processing 4) is performed on noise signals, and processing 5 is performed on copy signals and sine wave signals (that is, time envelope deformation processing is not performed). Is preferred.

(Modification 4 of the first embodiment)
The speech encoding device 11b (FIG. 44) of Modification 4 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory 11b is loaded into the RAM and executed to control the speech encoding device 11b in an integrated manner. The communication device of the audio encoding device 11b receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 11b includes a linear prediction analysis unit 1e1 instead of the linear prediction analysis unit 1e of the speech encoding device 11, and further includes a time slot selection unit 1p.

  The time slot selection unit 1p receives a signal in the QMF region from the frequency conversion unit 1a, and selects a time slot on which the linear prediction analysis processing in the linear prediction analysis unit 1e1 is performed. Based on the selection result notified from the time slot selection unit 1p, the linear prediction analysis unit 1e1 performs linear prediction analysis on the QMF region signal of the selected time slot in the same manner as the linear prediction analysis unit 1e, and performs a high-frequency linear prediction coefficient, low At least one of the frequency linear prediction coefficients is acquired. The filter strength parameter calculation unit 1f calculates the filter strength parameter using the linear prediction coefficient of the time slot selected by the time slot selection unit 1p obtained by the linear prediction analysis unit 1e1. In the selection of the time slot by the time slot selection unit 1p, for example, at least of the selection methods using the signal power of the QMF domain signal of the high frequency component similar to the time slot selection unit 3a in the decoding device 21a of the present modification described later. One may be used. At this time, the high-frequency component QMF domain signal in the time slot selector 1p is preferably a frequency component encoded by the SBR encoder 1d in the QMF domain signal received from the frequency converter 1a. As the time slot selection method, at least one of the above methods may be used, and at least one method different from the above method may be used, or a combination thereof may be used.

  The speech decoding device 21a (see FIG. 18) of Modification 4 of the first embodiment is physically provided with a CPU, ROM, RAM, communication device, etc. (not shown), and this CPU is a speech decoding device 21a such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 19) stored in the built-in memory is loaded into the RAM and executed, whereby the speech decoding apparatus 21a is controlled in an integrated manner. The communication device of the audio decoding device 21a receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 18, the speech decoding device 21a includes a low frequency linear prediction analysis unit 2d, a signal change detection unit 2e, a high frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, and a linear prediction filter. In place of the unit 2k, a low frequency linear prediction analysis unit 2d1, a signal change detection unit 2e1, a high frequency linear prediction analysis unit 2h1, a linear prediction inverse filter unit 2i1, and a linear prediction filter unit 2k3 are provided, and the time slot selection unit 3a is further provided. Prepare.

The time slot selection unit 3a performs linear prediction synthesis filter processing in the linear prediction filter unit 2k on the signal q _exp (k, r) of the QMF region of the high frequency component of the time slot r generated by the high frequency generation unit 2g. It is determined whether or not to perform, and a time slot for performing linear prediction synthesis filter processing is selected (processing in step Sh1). The time slot selection unit 3a notifies the selection result of the time slot to the low frequency linear prediction analysis unit 2d1, the signal change detection unit 2e1, the high frequency linear prediction analysis unit 2h1, the linear prediction inverse filter unit 2i1, and the linear prediction filter unit 2k3. . The low frequency linear prediction analysis unit 2d1 performs linear prediction analysis on the QMF region signal of the selected time slot r1 based on the selection result notified from the time slot selection unit 3a in the same manner as the low frequency linear prediction analysis unit 2d. A low frequency linear prediction coefficient is acquired (processing of step Sh2). Based on the selection result notified from the time slot selection unit 3a, the signal change detection unit 2e1 detects the time change of the QMF region signal in the selected time slot in the same manner as the signal change detection unit 2e, and the detection result T ( r1) is output.

The filter strength adjustment unit 2f performs filter strength adjustment on the low frequency linear prediction coefficient of the time slot selected by the time slot selection unit 3a obtained by the low frequency linear prediction analysis unit 2d1, and adjusts the linear prediction. The coefficient a _dec (n, r1) is obtained. The high-frequency linear prediction analysis unit 2h1 uses the high-frequency linear prediction analysis for the selected time slot r1 based on the selection result notified from the time slot selection unit 3a based on the QMF region signal of the high frequency component generated by the high frequency generation unit 2g. Similarly to the unit 2h, linear prediction analysis is performed in the frequency direction, and a high-frequency linear prediction coefficient a _exp (n, r1) is acquired (processing in step Sh3). Based on the selection result notified from the time slot selection unit 3a, the linear prediction inverse filter unit 2i1 converts the signal q _exp (k, r) of the high frequency component of the selected time slot r1 into the linear prediction inverse filter unit. Similar to 2i, linear prediction inverse filter processing is performed using a _exp (n, r1) as a coefficient in the frequency direction (processing of step Sh4).

ただし、Ｍは高周波生成部２ｇによって生成される高周波成分の下限周波数ｋ_ｘより高い周波数の範囲を表す値であり、さらには高周波生成部２ｇによって生成される高周波成分の周波数範囲をｋ_ｘ＜＝ｋ＜ｋ_ｘ＋Ｍのように表してもよい。また、所定の値Ｐ_{ｅｘｐ，Ｔｈ}は時間スロットｒを含む所定の時間幅のＰ_ｅｘｐ（ｒ）の平均値でもよい。さらに所定の時間幅はＳＢＲエンベロープでもよい。 In the linear prediction filter unit 2k3, based on the selection result notified from the time slot selection unit 3a, the signal _qadj (k, r1) in the QMF region of the high frequency component output from the high frequency adjustment unit 2j in the selected time slot r1. On the other hand, like the linear prediction filter unit 2k, linear prediction synthesis filter processing is performed in the frequency direction using a _adj (n, r1) obtained from the filter strength adjustment unit 2f (processing of step Sh5). Further, the change to the linear prediction filter unit 2k described in the modification 3 may be added to the linear prediction filter unit 2k3. In the selection of the time slot on which the linear prediction synthesis filter processing is performed in the time slot selection unit 3a, for example, a time slot in which the signal power of the high-frequency component QMF region signal q _exp (k, r) is larger than a predetermined value P _{exp, Th} One or more r may be selected. It is desirable to obtain the signal power of q _exp (k, r) by the following formula.

However, M is a value representing a frequency range higher than the lower limit frequency k _{x of the} high frequency component generated by the high frequency generation unit 2g, and further, the frequency range of the high frequency component generated by the high frequency generation unit 2g is represented by k _x ≦ It may be expressed as k <k _x + M. Further, the predetermined values P _{exp and Th} may be an average value of P _exp (r) having a predetermined time width including the time slot r. Further, the predetermined time width may be an SBR envelope.

について

が正の値から負の値に変わる時間スロットｒの高周波成分のＱＭＦ領域の信号電力をピークとしてもよい。信号電力の移動平均値

は、例えば次の式で求めることができる。

ただし、ｃは平均値を求める範囲を定める所定の値である。また信号電力のピークは、前記の方法で求めてもよく、異なる方法により求めてもよい。 Further, it may be selected so as to include a time slot in which the signal power of the high-frequency component QMF region signal reaches its peak. The peak of signal power is the moving average value of signal power

about

The signal power in the QMF region of the high-frequency component in the time slot r when the value changes from a positive value to a negative value may be peaked. Moving average value of signal power

Can be obtained by the following equation, for example.

However, c is a predetermined value that defines a range for obtaining an average value. The peak of signal power may be obtained by the above method or may be obtained by a different method.

Furthermore, the time width t from the steady state where the signal power fluctuation of the QMF region signal of the high frequency component is small to the transient state where the fluctuation is large is smaller than a predetermined value t _th , and the time slot included in the time width is at least One may be selected. Furthermore, the time width t from the transient state in which the signal power of the QMF region signal of the high frequency component is large to the steady state in which the variation is small is smaller than a predetermined value t _th, and at least the time slot included in the time width is at least One may be selected. A time slot r in which | P _exp (r + 1) −P _exp (r) | is smaller than (or equal to or smaller than a predetermined value) is set as the steady state, and | P _exp (r + 1) −P _exp ( r) | may be equal to or greater than (or greater than) a predetermined time slot r as the transient state, and | P _{exp, MA} (r + 1) −P _{exp, MA} (r) | A time slot r smaller than a predetermined value (or equal to or smaller than a predetermined value) is set as the steady state, and | P _{exp, MA} (r + 1) −P _{exp, MA} (r) | is equal to a predetermined value or A time slot r that is large (or larger than a predetermined value) may be the transient state. Further, the transient state and the steady state may be defined by the above-described method, or may be defined by different methods. As the time slot selection method, at least one of the above methods may be used, and at least one method different from the above method may be used, or a combination thereof may be used.

(Modification 5 of the first embodiment)
A speech encoding device 11c (FIG. 45) of Modification 5 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory 11c is loaded into the RAM and executed, thereby controlling the speech encoding device 11c in an integrated manner. The communication device of the audio encoding device 11c receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 11c includes a time slot selecting unit 1p1 and a bit stream multiplexing unit 1g4 in place of the time slot selecting unit 1p and the bit stream multiplexing unit 1g of the speech encoding device 11b of Modification 4.

  The time slot selection unit 1p1 selects a time slot similarly to the time slot selection unit 1p described in the fourth modification of the first embodiment, and sends the time slot selection information to the bit stream multiplexing unit 1g4. The bit stream multiplexing unit 1g4 includes the encoded bit stream calculated by the core codec encoding unit 1c, the SBR auxiliary information calculated by the SBR encoding unit 1d, and the filter strength calculated by the filter strength parameter calculation unit 1f. Are multiplexed with the time slot selection information received from the time slot selection unit 1p1, and the multiplexed bit stream is passed through the communication device of the speech encoding device 11c. Output. The time slot selection information is time slot selection information received by the time slot selection unit 3a1 in the speech decoding device 21b described later, and may include, for example, an index r1 of the time slot to be selected. Furthermore, for example, parameters used in the time slot selection method of the time slot selection unit 3a1 may be used. The speech decoding device 21b (see FIG. 20) of Modification 5 of the first embodiment is physically provided with a CPU, ROM, RAM, communication device, etc. (not shown), and this CPU is a speech decoding device 21b such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 21) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding apparatus 21b in an integrated manner. The communication device of the audio decoding device 21b receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside.

As shown in FIG. 20, the speech decoding device 21b replaces the bit stream separation unit 2a and the time slot selection unit 3a of the speech decoding device 21a of the fourth modification with a bit stream separation unit 2a5 and a time slot selection unit 3a1. The time slot selection information is input to the time slot selection unit 3a1. Similarly to the bit stream separation unit 2a, the bit stream separation unit 2a5 separates the multiplexed bit stream into filter strength parameters, SBR auxiliary information, and encoded bit stream, and further separates time slot selection information. The time slot selection unit 3a1 selects a time slot based on the time slot selection information sent from the bitstream separation unit 2a5 (processing in step Si1). The time slot selection information is information used for time slot selection, and may include, for example, an index r1 of the time slot to be selected. Further, for example, parameters used in the time slot selection method described in the fourth modification may be used. In this case, in addition to the time slot selection information, a high frequency component QMF region signal generated by the high frequency generation unit 2g is also input to the time slot selection unit 3a1. The parameter may be a predetermined value (for example, P _{exp, Th} , t _Th, etc.) used for selecting the time slot.

(Modification 6 of the first embodiment)
A speech encoding device 11d (not shown) of Modification 6 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory 11d is loaded into the RAM and executed to control the speech encoding device 11d in an integrated manner. The communication device of the audio encoding device 11d receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 11d includes a short-time power calculation unit 1i1 (not shown) and a time slot selection unit 1p2 instead of the short-time power calculation unit 1i of the speech encoding device 11a of the first modification.

  The time slot selection unit 1p2 receives a signal in the QMF region from the frequency conversion unit 1a, and selects a time slot corresponding to a time interval on which the short time power calculation unit 1i performs the short time power calculation process. Based on the selection result notified from the time slot selecting unit 1p2, the short time power calculating unit 1i1 converts the short time power of the time section corresponding to the selected time slot to the short time power of the speech encoding device 11a of the first modification. Calculation is performed in the same manner as the power calculation unit 1i.

(Modification 7 of the first embodiment)
A speech encoding device 11e (not shown) of Modification 7 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory of 11e is loaded into the RAM and executed to control the speech encoding device 11e in an integrated manner. The communication device of the audio encoding device 11e receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 11e includes a time slot selecting unit 1p3 (not shown) instead of the time slot selecting unit 1p2 of the speech encoding device 11d of the modification 6. Further, in place of the bit stream multiplexing unit 1g1, a bit stream multiplexing unit that further receives an output from the time slot selection unit 1p3 is provided. The time slot selection unit 1p3 selects a time slot similarly to the time slot selection unit 1p2 described in the sixth modification of the first embodiment, and sends the time slot selection information to the bit stream multiplexing unit.

(Modification 8 of the first embodiment)
A speech encoding apparatus (not shown) of Modification 8 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not physically shown, and this CPU is a speech of Modification 8 of ROM or the like. A predetermined computer program stored in the internal memory of the encoding device is loaded into the RAM and executed, whereby the speech encoding device of the modification 8 is controlled in an integrated manner. The communication device of the audio encoding device according to the modified example 8 receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding apparatus according to the modified example 8 further includes a time slot selecting unit 1p in addition to the speech encoding apparatus according to the modified example 2.

  The speech decoding apparatus (not shown) of Modification 8 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not physically shown, and this CPU performs speech decoding of Modification 8 of the ROM or the like. A predetermined computer program stored in the built-in memory of the apparatus is loaded into the RAM and executed, whereby the speech decoding apparatus of the modification 8 is comprehensively controlled. The communication device of the audio decoding device according to the modified example 8 receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. The speech decoding apparatus according to Modification 8 includes a low-frequency linear prediction analysis unit 2d, a signal change detection unit 2e, a high-frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, and a linear configuration of the speech decoding apparatus according to Modification 2. Instead of the prediction filter unit 2k, a low frequency linear prediction analysis unit 2d1, a signal change detection unit 2e1, a high frequency linear prediction analysis unit 2h1, a linear prediction inverse filter unit 2i1, and a linear prediction filter unit 2k3 are provided, and a time slot selection unit 3a Is further provided.

(Modification 9 of the first embodiment)
The speech encoding apparatus (not shown) of Modification 9 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically shown. This CPU is a speech of Modification 9 such as ROM. A predetermined computer program stored in the internal memory of the encoding device is loaded into the RAM and executed, whereby the speech encoding device of the modification 9 is controlled in an integrated manner. The communication device of the audio encoding device according to the modified example 9 receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding apparatus of Modification 9 includes a time slot selection unit 1p1 instead of the time slot selection unit 1p of the speech encoding apparatus described in Modification 8. Further, in place of the bit stream multiplexing unit described in the modification 8, in addition to the input to the bit stream multiplexing unit described in the modification 8, the bit stream multiplexing unit that further receives the output from the time slot selection unit 1p1 Is provided.

The speech decoding apparatus (not shown) of Modification 9 of the first embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not physically shown, and this CPU is a speech decoding of Modification 9 such as ROM. A predetermined computer program stored in the built-in memory of the apparatus is loaded into the RAM and executed, whereby the speech decoding apparatus of the modified example 9 is comprehensively controlled. The communication device of the audio decoding device according to the modified example 9 receives the encoded multiplexed bit stream and further outputs the decoded audio signal to the outside. The speech decoding apparatus according to Modification 9 includes a time slot selection unit 3a1 instead of the time slot selection unit 3a of the speech decoding apparatus according to Modification 8. Further, in place of the bit stream separation unit 2a, a bit stream separation unit is provided that separates a _D (n, r) described in the second modification in place of the filter strength parameter of the bit stream separation unit 2a5.

(Modification 1 of 2nd Embodiment)
The speech encoding device 12a (FIG. 46) according to the first modification of the second embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory 12a is loaded into the RAM and executed, thereby controlling the speech encoding device 12a in an integrated manner. The communication device of the audio encoding device 12a receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 12a includes a linear prediction analysis unit 1e1 instead of the linear prediction analysis unit 1e of the speech encoding device 12, and further includes a time slot selection unit 1p.

  The speech decoding device 22a (see FIG. 22) according to the first modification of the second embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated. The CPU includes a speech decoding device 22a such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 23) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding apparatus 22a in an integrated manner. The communication device of the audio decoding device 22a receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 22, the speech decoding device 22a includes a high-frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, a linear prediction filter unit 2k1, and a linear prediction interpolation / external device of the speech decoding device 22 according to the second embodiment. Instead of 2p, a high-frequency linear prediction analysis unit 2h1, a linear prediction inverse filter unit 2i1, a linear prediction filter unit 2k2, and a linear prediction interpolation / complementary external 2p1 are provided, and a time slot selection unit 3a is further provided.

The time slot selection unit 3a notifies the selection result of the time slot to the high frequency linear prediction analysis unit 2h1, the linear prediction inverse filter unit 2i1, the linear prediction filter unit 2k2, and the linear prediction coefficient interpolation / complementary external 2p1. In the linear prediction coefficient interpolation / external external 2p1, based on the selection result notified from the time slot selecting unit 3a, a _H (n, n, corresponding to the time slot r1 which is the selected time slot and the linear prediction coefficient is not transmitted. r) is obtained by interpolation or extrapolation in the same manner as the linear prediction coefficient interpolation / extrapolation 2p (processing of step Sj1). In the linear prediction filter unit 2k2, based on the selection result notified from the time slot selection unit 3a, for the selected time slot r1, the linear prediction coefficient is applied to q _adj (n, r1) output from the high frequency adjustment unit 2j. Using the interpolated or extrapolated a _H (n, r1) obtained from the interpolation / extrapolation 2p1, linear prediction synthesis filter processing is performed in the frequency direction in the same manner as the linear prediction filter unit 2k1 (in step Sj2). processing). Moreover, you may add the change to the linear prediction filter part 2k described in the modification 3 of 1st Embodiment to the linear prediction filter part 2k2.

(Modification 2 of the second embodiment)
The speech encoding device 12b (FIG. 47) of the second modification of the second embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory 12b is loaded into the RAM and executed to control the speech encoding device 11b in an integrated manner. The communication device of the audio encoding device 12b receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 12b includes a time slot selecting unit 1p1 and a bit stream multiplexing unit 1g5 in place of the time slot selecting unit 1p and the bit stream multiplexing unit 1g2 of the speech encoding device 12a of Modification 1. Similarly to the bit stream multiplexing unit 1g2, the bit stream multiplexing unit 1g5, the encoded bit stream calculated by the core codec encoding unit 1c, the SBR auxiliary information calculated by the SBR encoding unit 1d, and linear prediction A time slot index corresponding to the quantized linear prediction coefficient given from the coefficient quantization unit 1k is multiplexed, and further, time slot selection information received from the time slot selection unit 1p1 is multiplexed into a bit stream, and multiplexed bits The stream is output via the communication device of the audio encoding device 12b.

The speech decoding device 22b (see FIG. 24) of Modification 2 of the second embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not shown physically, and this CPU is a speech decoding device 22b such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 25) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding apparatus 22b in an integrated manner. The communication device of the audio decoding device 22b receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 24, the audio decoding device 22b replaces the bit stream separation unit 2a1 and the time slot selection unit 3a of the audio decoding device 22a described in the first modification with the bit stream separation unit 2a6 and the time slot selection. Time slot selection information is input to the time slot selection unit 3a1. In the bit stream separation unit 2a6, as in the bit stream separation unit 2a1, the multiplexed bit stream is quantized using a _H (n, r _i ), the index r _i of the corresponding time slot, and the SBR auxiliary Information and encoded bitstream are separated, and time slot selection information is further separated.

は、ｅ（ｒ）のＳＢＲエンベロープ内での平均値であってもよく、さらに別に定める値であってもよい。 (Modification 4 of the third embodiment) As described in Modification 1 of the third embodiment

May be an average value of e (r) within the SBR envelope, or may be a value determined separately.

(Modification 5 of the third embodiment)
As described in Modification 3 of the third embodiment, the envelope shape adjusting unit 2s has the adjusted time envelope e _adj (r) as expressed by, for example, Expression (28), Expression (37), and (38), In view of the fact that the gain coefficient is multiplied by the QMF subband sample, it is desirable to limit e _adj (r) by a predetermined value e _{adj, Th} (r) as follows.

(Fourth embodiment)
The speech encoding device 14 (FIG. 48) of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a built-in memory of the speech encoding device 14 such as a ROM. The voice encoding device 14 is centrally controlled by loading a predetermined computer program stored in the RAM into the RAM and executing it. The communication device of the audio encoding device 14 receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 14 includes a bit stream multiplexing unit 1g7 instead of the bit stream multiplexing unit 1g of the speech encoding device 11b according to the fourth modification of the first embodiment, and further includes the time of the speech encoding device 13. An envelope calculation unit 1m and an envelope shape parameter calculation unit 1n are provided.

  Similarly to the bit stream multiplexing unit 1g, the bit stream multiplexing unit 1g7 multiplexes the encoded bit stream calculated by the core codec encoding unit 1c and the SBR auxiliary information calculated by the SBR encoding unit 1d. Further, the filter strength parameter calculated by the filter strength parameter calculation unit and the envelope shape parameter calculated by the envelope shape parameter calculation unit 1n are converted into time envelope auxiliary information and multiplexed, and a multiplexed bit stream (encoding) is performed. The multiplexed bit stream) is output via the communication device of the audio encoding device 14.

(Modification 4 of the fourth embodiment)
The speech encoding device 14a (FIG. 49) according to the fourth modification of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory 14a is loaded into the RAM and executed to control the speech encoding device 14a in an integrated manner. The communication device of the audio encoding device 14a receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 14a includes a linear prediction analysis unit 1e1 instead of the linear prediction analysis unit 1e of the speech encoding device 14 of the fourth embodiment, and further includes a time slot selection unit 1p.

  A speech decoding device 24d (see FIG. 26) of Modification 4 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not shown physically, and this CPU is a speech decoding device 24d such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 27) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24d in an integrated manner. The communication device of the audio decoding device 24d receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. The speech decoding device 24d includes a low frequency linear prediction analysis unit 2d, a signal change detection unit 2e, a high frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, and a linear prediction filter as shown in FIG. In place of the unit 2k, a low frequency linear prediction analysis unit 2d1, a signal change detection unit 2e1, a high frequency linear prediction analysis unit 2h1, a linear prediction inverse filter unit 2i1, and a linear prediction filter unit 2k3 are provided, and the time slot selection unit 3a is further provided. Prepare. The temporal envelope deforming unit 2v uses the signal of the QMF region obtained from the linear prediction filter unit 2k3, the temporal envelope information obtained from the envelope shape adjusting unit 2s, as the third embodiment, the fourth embodiment, And it deform | transforms similarly to the time envelope deformation | transformation part 2v of those modifications (process of step Sk1).

(Modification 5 of the fourth embodiment)
The speech decoding device 24e (see FIG. 28) of Modification 5 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech decoding device 24e such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 29) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24e in an integrated manner. The communication device of the audio decoding device 24e receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 28, the speech decoding device 24 e of the speech decoding device 24 d according to Modification 4 can be omitted throughout the fourth embodiment in Modification 5 as in the first embodiment. The high frequency linear prediction analysis unit 2h1 and the linear prediction inverse filter unit 2i1 are omitted, and instead of the time slot selection unit 3a and the time envelope transformation unit 2v of the speech decoding device 24d, a time slot selection unit 3a2 and a time envelope transformation unit 2v1. Furthermore, the order of the linear prediction synthesis filter processing of the linear prediction filter unit 2k3 and the time envelope deformation processing in the time envelope deformation unit 2v1 that can change the processing order throughout the fourth embodiment are interchanged.

でのそれらの平均値である数式（２４）の

もあわせて時間スロット選択情報としてもよい。ただし、

である。 Similarly to the time envelope deformation unit 2v, the time envelope deformation unit 2v1 deforms q _adj (k, r) obtained from the high frequency adjustment unit 2j by using e _adj (r) obtained from the envelope shape adjustment unit 2s. Then, a signal _qenvadj (k, r) in the QMF region in which the time envelope is deformed is acquired. Further, the time slot selection unit 3a2 is notified of the parameters obtained during the time envelope transformation process or at least the parameters calculated using the parameters obtained during the time envelope transformation process as time slot selection information. The time slot selection information may be e (r) in Equation (22) or Equation (40) or | e (r) | ^{2 in} which the square root operation is not performed in the calculation process, and a plurality of time slot intervals (for example, SBR). envelope)

The average value of them in (24)

In addition, time slot selection information may be used. However,

It is.

でのそれらの平均値である

もあわせて時間スロット選択情報としてもよい。ただし、

である。さらに時間スロット選択情報としては、数式（２３）、数式（３５）、数式（３６）のｅ_ａｄｊ（ｒ）またはその算出過程にて平方根演算をしない｜ｅ_ａｄｊ（ｒ）｜^２でもよく、さらにある複数時間スロット区間（例えばＳＢＲエンベロープ）

でのそれらの平均値である

もあわせて時間スロット選択情報としてもよい。ただし、

である。さらに時間スロット選択情報としては、数式（３７）のｅ_{ａｄｊ，ｓｃａｌｅｄ}（ｒ）またはその算出過程にて平方根演算をしない｜ｅ_{ａｄｊ，ｓｃａｌｅｄ}（ｒ）｜^２でもよく、さらにある複数時間スロット区間（例えばＳＢＲエンベロープ）

でのそれらの平均値である

もあわせて時間スロット選択情報としてもよい。ただし、

である。さらに時間スロット選択情報としては、時間エンベロープが変形された高周波成分に対応するＱＭＦ領域信号の時間スロットｒの信号電力Ｐ_{ｅｎｖａｄｊ}（ｒ）またはそれの平方根演算をした信号振幅値

でもよく、さらにある複数時間スロット区間（例えばＳＢＲエンベロープ）

でのそれらの平均値である

もあわせて時間スロット選択情報としてもよい。ただし、

である。ただし、Ｍは高周波生成部２ｇによって生成される高周波成分の下限周波数ｋ_ｘより高い周波数の範囲を表す値であり、さらには高周波生成部２ｇによって生成される高周波成分の周波数範囲をｋ_ｘ≦ｋ＜ｋ_ｘ＋Mのように表してもよい。 Further, the time slot selection information may be e _exp (r) in Equation (26) or Equation (41) or | e _exp (r) | ^{2 in} which the square root operation is not performed in the calculation process, and a plurality of time slot intervals. (Eg SBR envelope)

Is their average value at

In addition, time slot selection information may be used. However,

It is. The more time slot selection information, Equation (23), equation (35), not the square root operation in _{e adj} (r) or a calculation process of Equation _{(36) | e adj (r} ) | 2 even better, further Some time slot intervals (eg SBR envelope)

Is their average value at

In addition, time slot selection information may be used. However,

It is. Further, the time slot selection information may be e _{adj, scaled} (r) in Expression (37) or | e _{adj, scaled} (r) | ^{2 in} which a square root operation is not performed in the calculation process, and a plurality of time slot intervals ( Eg SBR envelope)

Is their average value at

In addition, time slot selection information may be used. However,

It is. Further, as the time slot selection information, the signal power _{value Penvadj} (r) of the time slot r of the QMF domain signal corresponding to the high frequency component whose time envelope is deformed or the signal amplitude value obtained by calculating the square root thereof.

There may be more than one multiple time slot interval (eg SBR envelope)

Is their average value at

In addition, time slot selection information may be used. However,

It is. However, M is a value representing a frequency range higher than the lower limit frequency k _{x of the} high frequency component generated by the high frequency generation unit 2g, and further, the frequency range of the high frequency component generated by the high frequency generation unit 2g is represented by k _x ≦ k. It may be expressed as <k _x + M.

Based on the time slot selection information notified from the time envelope deforming unit 2v1, the time slot selecting unit 3a2 receives the signal _qenvadj of the high frequency component of the time slot r whose time envelope has been deformed by the time envelope deforming unit 2v1. For (k, r), it is determined whether or not the linear prediction synthesis filter processing is performed in the linear prediction filter unit 2k, and a time slot on which the linear prediction synthesis filter processing is performed is selected (processing in step Sp1).

のうち少なくともひとつを含んでいてもよく、ｕ_Ｔｈは、上記

のうち少なくともひとつを含んでもよい。またｕ_Ｔｈは、時間スロットｒを含む所定の時間幅（例えばＳＢＲエンベロープ）のｕ（ｒ）の平均値でもよい。さらに、ｕ（ｒ）がピークになる時間スロットが含まれるように選択してもよい。ｕ（ｒ）のピークは、前記第１の実施形態の変形例４における高周波成分のＱＭＦ領域信号の信号電力のピークの算出と同様に算出できる。さらに、前記第１の実施形態の変形例４における定常状態と過渡状態を、ｕ（ｒ）を用いて前記第１の実施形態の変形例４と同様に判断し、それに基づいて時間スロットを選択してもよい。時間スロットの選択方法は、前記の方法を少なくともひとつ用いてもよく、さらには前記とは異なる方法を少なくともひとつ用いてもよく、さらにはそれらを組み合わせてもよい。 In the time slot selection in the time slot selection unit 3a2 in the present modification, the parameter u (r) included in the time slot selection information notified from the time envelope modification unit 2v1 is a predetermined value u. One or more time slots r greater than _Th may be selected, and one or more time slots r for which u (r) is greater than or equal to a predetermined value u _Th may be selected. u (r) is the above e (r), | e (r) | ² , e _exp (r), | e _exp (r) | ² , e _adj (r), | e _adj (r) | ² , e _{adj, scaled} (r), | e _{adj, scaled} (r) | ² , P _envelope (r), and

At least one of the above, u _Th may be

May include at least one of them. U _Th may be an average value of u (r) in a predetermined time width (for example, SBR envelope) including the time slot r. Further, it may be selected to include a time slot in which u (r) peaks. The peak of u (r) can be calculated in the same manner as the calculation of the peak of the signal power of the QMF region signal of the high frequency component in the fourth modification of the first embodiment. Further, the steady state and the transient state in the modification 4 of the first embodiment are determined in the same manner as in the modification 4 of the first embodiment using u (r), and the time slot is selected based on the determination. May be. As the time slot selection method, at least one of the above methods may be used, and at least one method different from the above method may be used, or a combination thereof may be used.

(Modification 6 of 4th Embodiment)
A speech decoding device 24f (see FIG. 30) of Modification 6 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech decoding device 24f such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 29) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24f in an integrated manner. The communication device of the audio decoding device 24f receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 30, the speech decoding device 24 f of the speech decoding device 24 d according to the modification 4 can be omitted throughout the fourth embodiment in the modification 6 as in the first embodiment. The signal change detection unit 2e1, the high-frequency linear prediction analysis unit 2h1, and the linear prediction inverse filter unit 2i1 are omitted, and the time slot selection unit is replaced with the time slot selection unit 3a and the time envelope modification unit 2v of the speech decoding device 24d. 3a2 and a time envelope deformation unit 2v1. Furthermore, the order of the linear prediction synthesis filter processing of the linear prediction filter unit 2k3 and the time envelope deformation processing in the time envelope deformation unit 2v1 that can change the processing order throughout the fourth embodiment are interchanged.

Based on the time slot selection information notified from the time envelope deforming unit 2v1, the time slot selecting unit 3a2 receives the signal _qenvadj of the high frequency component of the time slot r whose time envelope has been deformed by the time envelope deforming unit 2v1. For (k, r), it is determined whether or not the linear prediction synthesis filter processing is performed in the linear prediction filter unit 2k3, a time slot on which the linear prediction synthesis filter processing is performed is selected, and the selected time slot is a low frequency linearity. Notify the prediction analysis unit 2d1 and the linear prediction filter unit 2k3.

(Modification 7 of the fourth embodiment)
The speech encoding device 14b (FIG. 50) of Modification 7 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated. The CPU is a speech encoding device such as a ROM. A predetermined computer program stored in the built-in memory 14b is loaded into the RAM and executed to control the speech encoding device 14b in an integrated manner. The communication device of the audio encoding device 14b receives an audio signal to be encoded from the outside, and further outputs an encoded multiplexed bit stream to the outside. The speech encoding device 14b includes a bit stream multiplexing unit 1g6 and a time slot selecting unit 1p1 instead of the bit stream multiplexing unit 1g7 and the time slot selecting unit 1p of the speech encoding device 14a of the fourth modification.

  Similarly to the bit stream multiplexing unit 1g7, the bit stream multiplexing unit 1g6, the encoded bit stream calculated by the core codec encoding unit 1c, the SBR auxiliary information calculated by the SBR encoding unit 1d, and the filter strength The time slot selection information received from the time slot selection unit 1p1 is multiplexed by multiplexing the filter strength parameter calculated by the parameter calculation unit and the time envelope auxiliary information obtained by converting the envelope shape parameter calculated by the envelope shape parameter calculation unit 1n. Are multiplexed, and a multiplexed bit stream (encoded multiplexed bit stream) is output via the communication device of the audio encoding device 14b.

  A speech decoding device 24g (see FIG. 31) of Modification 7 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not shown physically, and this CPU is a speech decoding device 24g such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 32) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24g in an integrated manner. The communication device of the audio decoding device 24g receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 31, the audio decoding device 24g replaces the bit stream separation unit 2a3 and the time slot selection unit 3a of the audio decoding device 24d described in Modification 4 with a bit stream separation unit 2a7 and a time slot selection unit. 3a1 is provided.

  The bit stream separation unit 2a7, like the bit stream separation unit 2a3, converts the time envelope auxiliary information, the SBR auxiliary information, and the encoded bit stream from the multiplexed bit stream input via the communication device of the audio decoding device 24g. And time slot selection information.

(Modification 8 of the fourth embodiment)
The speech decoding device 24h (see FIG. 33) of Modification 8 of the fourth embodiment is physically provided with a CPU, ROM, RAM, communication device, etc. (not shown), and this CPU is a speech decoding device 24h such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 34) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24h in an integrated manner. The communication device of the audio decoding device 24h receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 33, the speech decoding device 24h includes a low frequency linear prediction analysis unit 2d, a signal change detection unit 2e, a high frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, and In place of the linear prediction filter unit 2k, a low frequency linear prediction analysis unit 2d1, a signal change detection unit 2e1, a high frequency linear prediction analysis unit 2h1, a linear prediction inverse filter unit 2i1, and a linear prediction filter unit 2k3 are provided, and a time slot selection unit 3a is further provided. The primary high-frequency adjusting unit 2j1 is one or more of the processes in the “HF Adjustment” step in the SBR of the “MPEG-4 AAC”, similarly to the primary high-frequency adjusting unit 2j1 in the second modification of the fourth embodiment. (Step Sm1 processing). Similarly to the secondary high frequency adjustment unit 2j2 in the second modification of the fourth embodiment, the secondary high frequency adjustment unit 2j2 is one of the processes in the “HF Adjustment” step in the SBR of the “MPEG-4 AAC”. One or more processes are performed (the process of step Sm2). The processing performed by the secondary high-frequency adjusting unit 2j2 is preferably a process that is not performed by the primary high-frequency adjusting unit 2j1 among the processes in the “HF Adjustment” step in the SBR of “MPEG-4 AAC”.

(Modification 9 of the fourth embodiment)
A speech decoding device 24i (see FIG. 35) of Modification 9 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not shown physically, and this CPU is a speech decoding device 24i such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 36) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24i in an integrated manner. The communication device of the audio decoding device 24i receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 35, the speech decoding device 24i can be omitted throughout the fourth embodiment as in the first embodiment, and the high-frequency linear prediction analysis unit 2h1 of the speech decoding device 24h according to the modified example 8, The linear predictive inverse filter unit 2i1 is omitted, and a time envelope deforming unit 2v1 and a time slot selecting unit 3a2 are provided instead of the time envelope deforming unit 2v and the time slot selecting unit 3a of the speech decoding device 24h according to the modified example 8. Furthermore, the order of the linear prediction synthesis filter processing of the linear prediction filter unit 2k3 and the time envelope deformation processing in the time envelope deformation unit 2v1 that can change the processing order throughout the fourth embodiment are interchanged.

(Modification 10 of the fourth embodiment)
The speech decoding device 24j (see FIG. 37) of Modification 10 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not physically illustrated, and this CPU is a speech decoding device 24j such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 36) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24j in an integrated manner. The communication device of the audio decoding device 24j receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 37, the speech decoding device 24j can be omitted throughout the fourth embodiment as in the first embodiment. The signal change detection unit 2e1 of the speech decoding device 24h according to the modified example 8, the high-frequency linearity can be omitted. The prediction analysis unit 2h1 and the linear prediction inverse filter unit 2i1 are omitted, and the time envelope modification unit 2v1 and the time slot are replaced with the time envelope modification unit 2v and the time slot selection unit 3a of the speech decoding device 24h according to the modification 8. A selection unit 3a2 is provided. Furthermore, the order of the linear prediction synthesis filter processing of the linear prediction filter unit 2k3 and the time envelope deformation processing in the time envelope deformation unit 2v1 that can change the processing order throughout the fourth embodiment are interchanged.

(Modification 11 of the fourth embodiment)
A speech decoding device 24k (see FIG. 38) of Modification 11 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech decoding device 24k such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 39) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24k in an integrated manner. The communication device of the audio decoding device 24k receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 38, the audio decoding device 24k replaces the bit stream separation unit 2a3 and the time slot selection unit 3a of the audio decoding device 24h of Modification 8 with a bit stream separation unit 2a7 and a time slot selection unit 3a1. Prepare.

(Modification 12 of the fourth embodiment)
The speech decoding device 24q (see FIG. 40) of Modification 12 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like which are not physically illustrated, and this CPU is a speech decoding device 24q such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 41) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24q in an integrated manner. The communication device of the audio decoding device 24q receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 40, the speech decoding device 24q includes a low frequency linear prediction analysis unit 2d, a signal change detection unit 2e, a high frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, and Instead of the individual signal component adjustment units 2z1, 2z2, and 2z3, the low frequency linear prediction analysis unit 2d1, the signal change detection unit 2e1, the high frequency linear prediction analysis unit 2h1, the linear prediction inverse filter unit 2i1, and the individual signal component adjustment unit 2z4. 2z5 and 2z6 (the individual signal component adjustment unit corresponds to time envelope transformation means), and further includes a time slot selection unit 3a.

  At least one of the individual signal component adjustment units 2z4, 2z5, and 2z6 relates to the signal component included in the output of the primary high frequency adjustment unit based on the selection result notified from the time slot selection unit 3a. The QMF region signal is processed in the same manner as the individual signal component adjustment units 2z1, 2z2, 2z3 (step Sn1 processing). The processing performed using the time slot selection information is processing including linear prediction synthesis filter processing in the frequency direction among the processing in the individual signal component adjustment units 2z1, 2z2, and 2z3 described in Modification 3 of the fourth embodiment. It is desirable to include at least one of them.

  The processing in the individual signal component adjustment units 2z4, 2z5, and 2z6 may be the same as the processing of the individual signal component adjustment units 2z1, 2z2, and 2z3 described in the third modification of the fourth embodiment. The signal component adjustment units 2z4, 2z5, and 2z6 may perform time envelope transformation on each of a plurality of signal components included in the output of the primary high frequency adjustment unit using different methods. (If all of the individual signal component adjustment units 2z4, 2z5, and 2z6 are not processed based on the selection result notified from the time slot selection unit 3a, this is equivalent to the third modification of the fourth embodiment of the present invention). .

  The time slot selection results notified from the time slot selection unit 3a to each of the individual signal component adjustment units 2z4, 2z5, and 2z6 do not necessarily have to be the same, and may be all or a part of them.

  Further, in FIG. 40, the time slot selection unit 3a notifies the individual signal component adjustment units 2z4, 2z5, and 2z6 of the selection result of the time slot, but the individual signal component adjustment units 2z4, 2z5, A plurality of time slot selectors may be provided for notifying the result of selecting different time slots for each or a part of 2z6. At that time, among the individual signal component adjustment units 2z4, 2z5, and 2z6, the process 4 described in Modification 3 of the fourth embodiment (envelope shape adjustment similar to the time envelope modification unit 2v with respect to the input signal) After performing a process of multiplying each QMF subband sample by a gain coefficient using the time envelope obtained from the unit 2s, the output signal is further filtered from the filter strength adjustment unit 2f similar to the linear prediction filter unit 2k. The time slot selection unit for the individual signal component adjustment unit that performs frequency direction linear prediction synthesis filter processing using the obtained linear prediction coefficient) receives the time slot selection information from the time envelope transformation unit, and performs time slot selection processing May be performed.

(Modification 13 of the fourth embodiment)
The speech decoding device 24m (see FIG. 42) of Modification 13 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not shown physically, and this CPU is a speech decoding device 24m such as a ROM. A predetermined computer program (for example, a computer program for performing the processing shown in the flowchart of FIG. 43) stored in the built-in memory is loaded into the RAM and executed to control the speech decoding device 24m in an integrated manner. The communication device of the audio decoding device 24m receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. As shown in FIG. 42, the audio decoding device 24m replaces the bit stream separation unit 2a3 and the time slot selection unit 3a of the audio decoding device 24q of Modification 12 with a bit stream separation unit 2a7 and a time slot selection unit 3a1. Prepare.

(Modification 14 of the fourth embodiment)
A speech decoding device 24n (not shown) of Modification 14 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not physically shown, and this CPU includes a speech decoding device 24n such as a ROM. A predetermined computer program stored in the built-in memory is loaded into the RAM and executed, thereby comprehensively controlling the speech decoding device 24n. The communication device of the audio decoding device 24n receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. The speech decoding device 24n functionally includes a low frequency linear prediction analysis unit 2d, a signal change detection unit 2e, a high frequency linear prediction analysis unit 2h, a linear prediction inverse filter unit 2i, and a linear configuration of the speech decoding device 24a of the first modification. Instead of the prediction filter unit 2k, a low frequency linear prediction analysis unit 2d1, a signal change detection unit 2e1, a high frequency linear prediction analysis unit 2h1, a linear prediction inverse filter unit 2i1, and a linear prediction filter unit 2k3 are provided, and a time slot selection unit 3a Is further provided.

(Modification 15 of the fourth embodiment)
A speech decoding device 24p (not shown) of Modification 15 of the fourth embodiment includes a CPU, a ROM, a RAM, a communication device, and the like that are not physically shown. The voice decoding device 24p is controlled in an integrated manner by loading a predetermined computer program stored in the built-in memory into the RAM and executing it. The communication device of the audio decoding device 24p receives the encoded multiplexed bit stream, and further outputs the decoded audio signal to the outside. The speech decoding device 24p functionally includes a time slot selection unit 3a1 instead of the time slot selection unit 3a of the speech decoding device 24n of the modification example 14. Further, a bit stream separation unit 2a8 (not shown) is provided instead of the bit stream separation unit 2a4.

  Similarly to the bit stream separation unit 2a4, the bit stream separation unit 2a8 separates the multiplexed bit stream into SBR auxiliary information and encoded bit stream, and further separates into time slot selection information.

  11, 11 a, 11 b, 11 c, 12, 12 a, 12 b, 13, 14, 14 a, 14 b... Speech encoding device, 1 a... Frequency converter, 1 b. SBR encoding unit, 1e, 1e1 ... linear prediction analysis unit, 1f ... filter strength parameter calculation unit, 1f1 ... filter strength parameter calculation unit, 1g, 1g1, 1g2, 1g3, 1g4, 1g5, 1g6, 1g7 ... bitstream multiplexing 1h: high frequency frequency inverse transform unit, 1i ... short time power calculation unit, 1j ... linear prediction coefficient thinning unit, 1k ... linear prediction coefficient quantization unit, 1m ... temporal envelope calculation unit, 1n ... envelope shape parameter calculation unit, 1p, 1p1... Time slot selection unit, 21, 22, 23, 24, 24b, 24c... Speech decoding device, 2a, 2a1, 2a2 , 2a3, 2a5, 2a6, 2a7 ... bit stream separating unit, 2b ... core codec decoding unit, 2c ... frequency converting unit, 2d, 2d1 ... low frequency linear prediction analysis unit, 2e, 2e1 ... signal change detecting unit, 2f ... filter Intensity adjustment unit, 2g ... high frequency generation unit, 2h, 2h1 ... high frequency linear prediction analysis unit, 2i, 2i1 ... linear prediction inverse filter unit, 2j, 2j1, 2j2, 2j3, 2j4 ... high frequency adjustment unit, 2k, 2k1, 2k2, 2k3 ... linear prediction filter unit, 2m ... coefficient addition unit, 2n ... frequency inverse transformation unit, 2p, 2p1 ... linear prediction coefficient interpolation / external, 2r ... low frequency time envelope calculation unit, 2s ... envelope shape adjustment unit, 2t ... High frequency time envelope calculation unit, 2u... Time envelope flattening unit, 2v, 2v1... Time envelope transformation unit, 2w. Z1,2z2,2z3,2z4,2z5,2z6 ... individual signal component adjuster, 3a, 3a1 and 3a2 ... time slot selection unit.

Claims (7)

An audio decoding device for decoding an encoded audio signal,
Bitstream separation means for separating an external bitstream including the encoded audio signal into an encoded bitstream and time envelope auxiliary information;
Core decoding means for decoding the encoded bitstream separated by the bitstream separation means to obtain a low frequency component;
Frequency converting means for converting the low frequency component obtained by the core decoding means into a frequency domain;
High frequency generation means for generating a high frequency component by copying the low frequency component converted into the frequency domain by the frequency conversion means from a low frequency band to a high frequency band;
Primary high-frequency adjusting means for generating an output signal by executing a part of processing including gain adjustment, noise superimposition, and sine wave addition processing on the high-frequency component generated by the high-frequency generating means When,
Low frequency time envelope analyzing means for analyzing the low frequency component converted into the frequency domain by the frequency converting means to obtain time envelope information;
Auxiliary information converting means for converting the time envelope auxiliary information into a parameter for adjusting the time envelope information;
Time envelope adjusting means for adjusting the time envelope information acquired by the low frequency time envelope analyzing means to generate adjusted time envelope information, and using the parameter for adjusting the time envelope information An envelope adjusting means;
Using the adjusted time envelope information, deforming the time envelope of the output signal generated by the primary high-frequency adjusting means, and generating an output signal;
Secondary high frequency adjustment means for performing other part of the processing including gain adjustment, noise superimposition, and sine wave addition processing on the output signal generated by the time envelope deformation means;
A speech decoding apparatus comprising: An audio decoding device for decoding an encoded audio signal,
Core decoding means for decoding a bitstream from the outside including the encoded audio signal to obtain a low frequency component;
Frequency converting means for converting the low frequency component obtained by the core decoding means into a frequency domain;
High frequency generation means for generating a high frequency component by copying the low frequency component converted into the frequency domain by the frequency conversion means from a low frequency band to a high frequency band;
Primary high-frequency adjusting means for generating an output signal by executing a part of processing including gain adjustment, noise superimposition, and sine wave addition processing on the high-frequency component generated by the high-frequency generating means When,
Low frequency time envelope analyzing means for analyzing the low frequency component converted into the frequency domain by the frequency converting means to obtain time envelope information;
A time envelope auxiliary information generator for analyzing the bitstream and generating parameters for adjusting the time envelope information;
Time envelope adjusting means for adjusting the time envelope information acquired by the low frequency time envelope analyzing means to generate adjusted time envelope information, and using the parameter for adjusting the time envelope information An envelope adjusting means;
Using the adjusted time envelope information, deforming the time envelope of the output signal generated by the primary high-frequency adjusting means, and generating an output signal;
Secondary high frequency adjustment means for performing other part of the processing including gain adjustment, noise superimposition, and sine wave addition processing on the output signal generated by the time envelope deformation means;
A speech decoding apparatus comprising:   3. The speech decoding apparatus according to claim 1, wherein the secondary high-frequency adjusting unit performs the addition process of the sine wave in an SBR decoding process on the output signal generated by the time envelope deforming unit. . A speech decoding method using a speech decoding device that decodes an encoded speech signal,
A bitstream separation step in which the speech decoding apparatus separates an external bitstream including the encoded speech signal into an encoded bitstream and time envelope auxiliary information;
A core decoding step in which the speech decoding apparatus obtains a low-frequency component by decoding the encoded bitstream separated in the bitstream separation step;
A frequency conversion step in which the speech decoding apparatus converts the low frequency component obtained in the core decoding step into a frequency domain;
A high frequency generation step in which the speech decoding apparatus generates a high frequency component by copying the low frequency component converted into the frequency domain in the frequency conversion step from a low frequency band to a high frequency band;
The speech decoding apparatus generates an output signal by executing a part of processing including gain adjustment, noise superimposition, and sine wave addition processing on the high-frequency component generated in the high-frequency generation step. A primary high frequency adjustment step,
A low-frequency time envelope analysis step in which the speech decoding apparatus acquires time envelope information by analyzing the low-frequency component converted into the frequency domain in the frequency conversion step;
An auxiliary information converting step in which the speech decoding apparatus converts the time envelope auxiliary information into a parameter for adjusting the time envelope information;
The speech decoding apparatus is a time envelope adjustment step of adjusting the time envelope information acquired in the low frequency time envelope analysis step to generate adjusted time envelope information, and the parameter is used to adjust the time envelope information. Using the time envelope adjustment step;
The speech decoding apparatus uses the adjusted time envelope information to modify the time envelope of the output signal generated in the primary high frequency adjustment step, thereby generating an output signal;
The speech decoding apparatus performs secondary part of the processing including gain adjustment, noise superimposition, and sine wave addition processing on the output signal generated in the time envelope transformation step. A high frequency adjustment step;
A speech decoding method including: A speech decoding method using a speech decoding device that decodes an encoded speech signal,
A core decoding step in which the speech decoding apparatus obtains a low-frequency component by decoding an external bitstream including the encoded speech signal;
A frequency conversion step in which the speech decoding apparatus converts the low frequency component obtained in the core decoding step into a frequency domain;
A high frequency generation step in which the speech decoding apparatus generates a high frequency component by copying the low frequency component converted into the frequency domain in the frequency conversion step from a low frequency band to a high frequency band;
The speech decoding apparatus generates an output signal by executing a part of processing including gain adjustment, noise superimposition, and sine wave addition processing on the high-frequency component generated in the high-frequency generation step. A primary high frequency adjustment step,
A low-frequency time envelope analysis step in which the speech decoding device acquires time envelope information by analyzing the low-frequency component converted into the frequency domain in the frequency conversion step;
A time envelope auxiliary information generating step in which the speech decoding device generates a parameter for analyzing the bitstream and adjusting the time envelope information;
The speech decoding apparatus is a time envelope adjustment step of adjusting the time envelope information acquired in the low frequency time envelope analysis step to generate adjusted time envelope information, and the parameter is used to adjust the time envelope information. Using the time envelope adjustment step;
The speech decoding apparatus uses the adjusted time envelope information to modify the time envelope of the output signal generated in the primary high frequency adjustment step, thereby generating an output signal;
The speech decoding apparatus performs secondary part of the processing including gain adjustment, noise superimposition, and sine wave addition processing on the output signal generated in the time envelope transformation step. A high frequency adjustment step;
A speech decoding method including: In order to decode the encoded audio signal, a computer device is
Bitstream separation means for separating an external bitstream including the encoded audio signal into an encoded bitstream and time envelope auxiliary information;
Core decoding means for decoding the encoded bitstream separated by the bitstream separation means to obtain a low frequency component;
Frequency converting means for converting the low frequency component obtained by the core decoding means into a frequency domain;
High frequency generation means for generating a high frequency component by copying the low frequency component converted into the frequency domain by the frequency conversion means from a low frequency band to a high frequency band;
Primary high-frequency adjusting means for generating an output signal by executing a part of processing including gain adjustment, noise superimposition, and sine wave addition processing on the high-frequency component generated by the high-frequency generating means When,
Low frequency time envelope analyzing means for analyzing the low frequency component converted into the frequency domain by the frequency converting means to obtain time envelope information;
Auxiliary information converting means for converting the time envelope auxiliary information into a parameter for adjusting the time envelope information;
Time envelope adjusting means for adjusting the time envelope information acquired by the low frequency time envelope analyzing means to generate adjusted time envelope information, and using the parameter for adjusting the time envelope information An envelope adjusting means;
Using the adjusted time envelope information, deforming the time envelope of the output signal generated by the primary high-frequency adjusting means, and generating an output signal;
Secondary high frequency adjustment means for performing other part of the processing including gain adjustment, noise superimposition, and sine wave addition processing on the output signal generated by the time envelope deformation means;
Voice decoding program that functions as In order to decode the encoded audio signal, a computer device is
Core decoding means for decoding a bitstream from the outside including the encoded audio signal to obtain a low frequency component;
Frequency converting means for converting the low frequency component obtained by the core decoding means into a frequency domain;
High frequency generation means for generating a high frequency component by copying the low frequency component converted into the frequency domain by the frequency conversion means from a low frequency band to a high frequency band;
Primary high-frequency adjusting means for generating an output signal by executing a part of processing including gain adjustment, noise superimposition, and sine wave addition processing on the high-frequency component generated by the high-frequency generating means When,
Low frequency time envelope analyzing means for analyzing the low frequency component converted into the frequency domain by the frequency converting means to obtain time envelope information;
A time envelope auxiliary information generator for analyzing the bitstream and generating parameters for adjusting the time envelope information;
Time envelope adjusting means for adjusting the time envelope information acquired by the low frequency time envelope analyzing means to generate adjusted time envelope information, and using the parameter for adjusting the time envelope information An envelope adjusting means;
Using the adjusted time envelope information, deforming the time envelope of the output signal generated by the primary high-frequency adjusting means, and generating an output signal;
Secondary high frequency adjustment means for performing other part of the processing including gain adjustment, noise superimposition, and sine wave addition processing on the output signal generated by the time envelope deformation means;
Voice decoding program to function as.

Images