JP2002198870A - Echo processing device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an echo processing device in which a pseudo background noise with high quality is generated by a simple process. SOLUTION: The device is provided with a sound coding means for generating and inputting coded data equivalent to a background noise by using an analytic parameter for coding obtained in an analytic frame to be previously judged as a noise section in an analytic frame judged as a frame during which only a near end background noise component and an echo signal are contained in a transmission signal.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an echo processing for reducing an echo included in a transmission audio signal, which is generated in a communication path or an echo path between a speaker and a microphone, in voice communication such as a fixed telephone, a vehicle-mounted telephone, and a portable telephone. It concerns the device.

[0002]

2. Description of the Related Art A conventional echo processing apparatus is disclosed in
No. 17589 was disclosed. FIG. 7 is a simplified block diagram of a conventional echo processing apparatus disclosed in Japanese Patent Application Laid-Open No. 11-17589. For the sake of explanation, in addition to the echo processing apparatus main body, audio encoding means for encoding a transmission signal Voice decoding means for decoding the received signal, voice coding means for coding the voice signal at the far end, and voice decoding means for decoding the voice signal.

In FIG. 7, reference numeral 1 denotes a pseudo echo generating means,
2 and 7 are addition means, 3 is an echo canceller means, 4 and 5 are sound / silence determination means, 6 is echo suppression means, 8 is pseudo background noise generation means, and these constitute an echo processing apparatus main body. I have. Reference numeral 9 denotes an audio encoding unit, and reference numeral 10 denotes an audio decoding unit. These components constitute an audio encoding / decoding unit 11. Reference numeral 18 denotes a far-end speech decoding unit, and reference numeral 19 denotes a far-end speech encoding unit.

Next, the operation will be described. The speech signal uttered by the far end speaker is encoded by the speech encoding unit 19 and input to the near end speech decoding unit 10 as encoded data CR via a communication path. Voice decoding means 10
Decodes the encoded data CR and outputs a received signal RI (i) as an analog signal. The sound / silence determining means 5 determines whether the received signal RI (i) is sound or no sound based on the power of the received signal RI (i), and determines the result as the echo suppressor 6 and the pseudo background noise generating means 8. Output to
Also, the received signal RI (i) is directly received signal RO
(I), which is output to the outside including the echo path.

On the other hand, the transmission signal SI (i) input to the echo canceller means 3 includes a reception signal RO (i) converted into an analog signal and an echo signal input as an echo through an echo path. The speech signal uttered by the end speaker and the background noise at the near end are included. The echo canceller means 3 estimates the transfer characteristic of the echo path, generates a pseudo echo signal by the pseudo echo generation means 1, subtracts the pseudo echo signal from the received signal SI (i) by using the addition means 2, and generates a residual signal SA. Find (i). The voiced / silence determining means 4 determines whether the residual signal SA (i) is voiced or silent based on the power of the residual signal SA (i), and sends the determination result to the echo suppressor 6 and the pseudo background noise generating means 8. Output. The power is calculated by calculating the sum of squares of the samples.

The echo suppressor 6 receives the received signal RI
The section where (i) is determined to be sound by the sound / silence determination means 5 and the residual signal SA (i) is determined to be silent by the sound / silence determination means 4 is included in the residual signal SA (i). It is determined that the signal is only an echo signal, and the echo signal is suppressed by suppressing the amplitude of the residual signal SA (i). If both the determination result of the voiced / silent determination means 5 and the determination result of the voiced / silence determination means 4 are sound, or if the determination result of the voiced / silence determination means 5 is silent, the amplitude is SA (i). Is not suppressed.

[0007] The pseudo background noise generating means 8 receives the received signal RI.
(I) Calculates and stores the spectrum parameter (linear prediction coefficient) of the section determined to be silent by the voiced / silent determining means 5. Then, the echo suppressor 6 outputs the residual signal SA.
In the section in which the amplitude of (i) is suppressed, a synthesis filter process is performed using the spectrum parameters obtained in the silent section and white noise (referred to as white noise, which has a flat frequency spectrum shape). Generates pseudo background noise. This pseudo background noise is added to the signal whose amplitude has been suppressed by the echo suppressor 6 via the adder 7, and a transmission signal SO (i) is obtained from the adder 7.

By this processing, the background noise once suppressed together with the echo signal by the eco-suppressing means 6 is compensated for by the pseudo background noise having a spectral characteristic close to that of the actual background noise. The above discomfort is reduced. Next, the transmission signal SO (i) is encoded by the audio encoding means 9 and output as encoded data CS. The encoded data CS is input to the voice decoding means 18 on the far end side via the communication path. Voice decoding means 1
Numeral 8 decodes the encoded data and outputs an audio signal.

[0009]

Since the conventional echo processing apparatus is configured as described above, when generating a pseudo background noise, a process of calculating a spectrum parameter and a process of calculating a pseudo background from the spectrum parameter and white noise are performed. It is necessary to perform synthesis filter processing for generating noise, and there is a problem that the processing is complicated and the amount of calculation is large.

SUMMARY OF THE INVENTION The present invention has been made to solve the above-mentioned conventional problems, and has as its object to provide an echo processing apparatus which generates high-quality pseudo background noise by simple processing.

[0011]

An echo processing apparatus according to the present invention is characterized in that an echo determining means determines that only a near-end background noise component and a received signal are echoed by an echo path in a transmission signal. And the transmission signal is encoded, and in the analysis frame in which the echo determination means determines that only the near-end background noise component and the echo signal are present in the transmission signal, the analysis frame determined in advance is a noise section. Speech encoding means for generating and outputting encoded data corresponding to background noise using the encoded analysis parameters.

In the echo processing apparatus according to the present invention, the speech encoding method of the speech encoding means is the CELP method, and the encoded data corresponding to the background noise is a linear prediction coefficient, an adaptive excitation code, a driving excitation code, and both of them. It is composed of the gain of the excitation code.

In the echo processing apparatus according to the present invention, in the analysis frame in which only the near-end background noise component and the echo signal are present in the transmission signal in the echo determination means, a plurality of analysis frames previously determined to be noise sections are provided. A plurality of coded spectral parameters obtained in the analysis frame are randomly selected and sequentially output.

In the echo processing apparatus according to the present invention, in the analysis frame in which the echo determination means determines that only the near-end background noise component and the echo signal are present in the transmission signal, a plurality of the noise frames are determined in advance as noise sections. A plurality of spectral parameters obtained in the analysis frame are randomly selected, encoded, and sequentially output.

In the echo processing apparatus according to the present invention, in the analysis frame in which only the near-end background noise component and the echo signal are present in the transmission signal in the echo determination means, a plurality of noise sections previously determined as noise sections are provided. The spectrum parameters in the analysis frame are averaged, and the values of each dimension of the averaged spectrum parameter are fluctuated for each analysis frame, encoded, and sequentially output.

According to the echo processing apparatus of the present invention, the adaptive excitation code, the driving excitation code, and the coded gains corresponding to both excitation codes, which are obtained in the same analysis frame in which the selected spectral parameter is obtained, are obtained. And outputs them sequentially.

In the echo processing apparatus according to the present invention, in the analysis frame in which the echo determination means determines that only the near-end background noise component and the echo signal are present in the transmission signal, a plurality of the noise frames are determined in advance as noise sections. A plurality of adaptive excitation codes determined in the analysis frame and a set of encoded gains corresponding to the respective adaptive excitation codes are randomly selected and sequentially output, and also determined in a plurality of analysis frames previously determined to be noise sections. A plurality of drive excitation codes and a set of coded gains corresponding to the respective drive excitation codes are randomly selected and sequentially output.

In the echo processing apparatus according to the present invention, the gain of the adaptive excitation code and the gain of the driving excitation code are obtained and held before encoding, and are encoded and output each time they are selected.

The echo processing apparatus according to the present invention corrects the gains of the selected adaptive excitation code and driving excitation code based on the average power of the target signal obtained in advance in a plurality of analysis frames determined as noise sections. And outputs it.

In the echo processing apparatus according to the present invention, when the gain of the driving excitation code is obtained in an analysis frame which is determined to be a noise section in advance, the adaptive excitation code gain is set to a value between zero and a value close to zero. Is obtained.

In the echo processing apparatus according to the present invention, a DTX (Discontinuous Transaction) is used for an analysis frame in which only a near-end background noise component and an echo signal are present in the transmission signal in the echo determination means.
issue) and output the silence flag in
It is provided with a speech encoding unit that operates internally in a silent section processing mode in DTX.

The echo processing apparatus according to the present invention comprises a voice coding means for randomly selecting, averaging and coding a plurality of spectral parameters obtained in advance in a plurality of analysis frames determined to be noise sections, and sequentially outputting them. Things.

The echo processing apparatus according to the present invention includes an encoding unit. The encoding unit, based on a result of the determination by the echo determining unit, includes: The image processing apparatus further includes a background noise generation unit that outputs coded data corresponding to background noise stored in advance instead of coded data of the transmission signal as transmission data.

[0024]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Embodiment 1 FIG. 1 is a block diagram showing a configuration of an echo processing device according to the present invention.
FIG. 2 is a block diagram showing a detailed configuration of the audio encoding unit in FIG. The conventional echo processing apparatus has provided a pseudo-background noise generating means to perform processing for generating pseudo-background noise. However, in the echo processing apparatus according to the first embodiment shown in FIG. Is generated.

The first embodiment of the present invention will be described below with reference to FIGS. In FIG. 1, the same reference numerals as those shown in FIG. 7 indicate the same or corresponding parts, and therefore, description of the same operations will be omitted. Reference numeral 12 denotes an echo determination unit, and 13 denotes a voice encoding unit that inputs the determination result of the echo determination unit 12 and the determination result of the voiced / silence determination unit 4.

In FIG. 2 showing a specific configuration of the speech coding means 13, reference numeral 51 denotes an adaptive excitation codebook, 52 denotes a driving excitation codebook, 53 and 54 denote amplifiers, 55 denotes addition means, and 56 denotes linear prediction analysis. Means, 57 is a synthesis filter means, 58 is an addition means, 59 is a sound source search means, 60 is a background noise generation means, and 61 is a multiplexing means.

In FIG. 1, echo determination means 12 receives a received signal RI (i) output from speech decoding means 10,
Residual signal SA output from echo canceller means 3
(I) is input, and the respective powers are obtained as RP and SAP for each analysis frame. Then, for example, it is determined whether or not the condition shown in Expression (1) is satisfied.

RP> SAP and SAP <TH (1) RP and SAP are simply symbols. TH is a fixed threshold value set in advance. In the analysis frame satisfying the condition of the expression (1), it is determined that only the echo signal is included in the residual signal SA (i) (hereinafter, referred to as “echo residual frame”), and the determination result is expressed as a speech code. Output to the converting means 13.

The sound / non-speech determining means 4 outputs the residual signal SA (i)
Is a sound or a silence, for example, the residual signal SA
The power and spectrum of (i) are obtained and determined, and the result of the determination is output to the speech encoding unit 13. The power is calculated by calculating the sum of squares of the samples.

The speech encoding means 13 is constituted by a CELP encoding scheme which is very often used in a standard speech codec scheme for a cellular phone or the like as shown in FIG. 2, and converts the residual signal SA (i) for each analysis frame. In addition to encoding, the encoded data corresponding to the pseudo background noise is generated according to the voiced / silent determination result output from the voiced / non-voiced determination unit 4 and the echo determination result output from the echo determination unit 12.

The operation of the speech encoding means 13 for generating encoded data corresponding to pseudo background noise will be described in detail below with reference to FIG.

In FIG. 2, since the configuration other than the background noise generating means 60 is the same as the configuration of the coding section of the general CELP coding method, first, the operation other than the background noise generating means 60 (ie, the CELP coding method) ) Will be briefly described.

The linear prediction analysis means 56 outputs the residual signal SA
(I), a linear predictive analysis is performed for each analysis frame, and for example, LSP (Lin
e Spectrum Pair) is obtained and quantized and output to the synthesis filter means 57. Also, the quantized LS
P and encodes the background noise generation means 60 and the multiplexing means 61
Output to Adaptive excitation codebook 51 stores past excitation signals. The excitation signal in adaptive excitation codebook 51 is cut out at a length (variable length) called a lag, and the cut-out signal is repeated at a lag cycle until the subframe length is reached, to generate an adaptive excitation. Driving excitation codebook 52 is composed of a plurality of noise signal vectors.

The adaptive excitation of the adaptive excitation codebook 51 and the driving excitation of the driving excitation codebook 52 are sequentially read out in units of subframe lengths, amplified by amplifiers 53 and 54 and given gains, respectively. And a sound source signal is obtained. The synthesis filter means 57 is provided for the linear prediction analysis means 5.
The synthesized speech is synthesized by using the LSP from S.6 and the sound source signal from the adding means 55. The addition means 58 adds the synthesized speech synthesized by the synthesis filter means 57 and the residual signal SA (i) to obtain distortion. The sound source searching means 59 searches and encodes the lag of the adaptive sound source and the driving sound source code and the gain for them when the distortion is minimized, and outputs the lag to the background noise generating means 60 and the multiplexing means 61.

The multiplexing means 61 multiplexes these coded data and outputs them as coded data CS to the communication path.
The coded data CS is transmitted to the voice decoding means 18 on the far end via the communication path, and is decoded by the voice decoding means 18 to obtain a decoded signal.

The above is the operation of the CELP coding system.
Next, the operation of the background noise generating means 60 will be described. The background noise generation unit 60 receives, in addition to the encoded data described above, the voiced / silent determination result output from the voiced / silence determination unit 4 in FIG. 1 and the echo determination result output from the echo determination unit 12.

First, the background noise generating means 60 calculates the LSP previously obtained from the voiced / silent determining means 4 output from the voiced / silent determining means 4 in the analysis frame in which the voice is silent, for example, as shown in FIG.
As shown in (1), the last six frames are always stored. Also,
As shown in FIGS. 3 (b) and 3 (c), the adaptive sound source and the sign of the gain and the sign of the driving sound source and the gain obtained in the same analysis frame in which the result of the sound / silence determination is silence are the latest 12 as shown in FIGS.
Subframes (one analysis frame corresponds to two subframes) are always stored.

Then, in the analysis frame in which the determination result of the echo determination means 12 is an echo residual frame, the background noise generation means 60 performs the operation described below. At this time, the speech encoding process by the basic operation of the CELP encoding described above is not performed.

The background noise generating means 60 first randomly selects one of the stored six LSPs from the stored six LSPs using random numbers (see FIG. 3), and outputs it to the multiplexing means 61. Next, a set of an adaptive excitation code and its gain code and a drive excitation code and its gain code in two subframes corresponding to the analysis frame in which the selected LSP is analyzed are selected (FIG. 3B, (See the dotted line in (c)) and output to the multiplexing means 61.

The multiplexing means 61 sequentially outputs coded data obtained in the same analysis frame in a past silent section (background noise section), so that a decoded signal obtained by decoding is: It has frequency characteristics and power as actual background noise, and shows good quality. In addition, since the encoded data of different analysis frames are sequentially output at random without repeating the same, the decoded signal decoded by the decoding unit that receives the encoded data includes an inappropriate period as background noise. There is no sex.

The contents of the coded data CS output from the background noise generating means 60 are the same as the contents of the coded data CS described in the basic operation of the CELP coding. Therefore, the voice decoding means 18 on the far end that receives and decodes the coded data CS via the communication path,
Is generated by decoding the speech signal in the speech section and the background noise in the echo residual frame section by ordinary CELP decoding processing, regardless of whether or not is generated by the background noise generation means 60.

As described above, according to the first embodiment, in the echo residual frame, the speech encoding means 13 based on the CELP method uses the LSP, adaptive excitation code, Since the drive excitation code and the set of coded gains are randomly selected and sequentially output to generate coded data corresponding to background noise, a new means for generating pseudo background noise is provided. In addition, by using a simple configuration and method using the widely-used CELP method, a high-quality pseudo-background noise having no periodicity and a characteristic as a background noise is generated in a section where only an echo exists in a transmission signal. Can be.

Embodiment 2 The background noise generating means 60 according to the first embodiment selects the adaptive excitation code, the driving excitation code, and their gain based on the analysis frame from which the LSP has been selected. And a set of gains thereof may be individually and randomly selected.

As described above, according to the second embodiment, the LSP, the adaptive excitation code, the driving excitation code, and their gains are individually and randomly selected. The simulated background noise has an effect that the random noise property is further increased and does not have a biased characteristic.

Embodiment 3 The background noise generation unit 60 according to the first embodiment holds and outputs the encoded LSP and gain in advance, but retains and outputs the actual LSP or gain value before encoding. It may be encoded.

As described above, according to the third embodiment, since the actual LSP or gain value is retained and re-encoded, for example, the LSP (or gain) of the previous analysis frame and the LSP (current gain) of the current frame are Alternatively, there is an effect that it is possible to cope with a case of using a method of encoding the parameter by utilizing the association with another parameter such as encoding the difference of the gain).

Embodiment 4 In the background noise generating means 60 of the first embodiment, the gains of the adaptive excitation code and the driving excitation code previously obtained in the silent section are output as they are. But,
The excitation signal (in the adaptive excitation codebook) obtained in the analysis frame of the silent section is used as the target signal, the average value of the power is determined as the reference excitation power, and the adaptive excitation code and the driving excitation code selected by the background noise generation unit 60 are selected. Alternatively, the gain of both excitation codes may be corrected and coded so that the power of the excitation signal generated using the gain of both excitation codes matches the reference excitation power. Also, at this time, the gain of the driving excitation may be obtained by setting the gain of the adaptive excitation code to zero or a value close to zero so that the power of the excitation signal generated only by the driving excitation code matches the reference excitation power.

As described above, according to the fourth embodiment, since the power of the excitation signal generated by the encoded data generated by the background noise generation means 60 matches the reference excitation power, there is no stable power fluctuation. There is an effect that the generated pseudo background noise can be generated. In addition, since the adaptive excitation is not used by setting the gain of the adaptive excitation to zero, even if the content of the adaptive excitation codebook has the characteristic of a voiced section by mistake,
Good quality pseudo background noise can be generated regardless of the contents of the adaptive excitation codebook.

Embodiment 5 The background noise generation means 60 according to the fourth embodiment obtains a reference sound source power using the sound source signal in a silent section as a target signal, and
The power of the excitation signal generated by the coded data generated by 0 is set to match the reference excitation power. However, the residual signal SA (i) in the silent section is set as the target signal, and the average value of the power is used as the reference sound. The two-source excitation gain is calculated so that the power of the synthesized signal synthesized using the LSP selected by the background noise generation means 60, the adaptive excitation code, the driving excitation code, and the gain of both excitation codes matches the reference audio power. May be modified.

As described above, according to the fifth embodiment, since the power of the synthesized signal generated by the encoded data generated by the background noise generating means 60 matches the reference audio power, the power does not fluctuate and is stable. There is an effect that pseudo background noise can be generated.

Embodiment 6 FIG. FIG. 4 is a block diagram showing a configuration of an echo processing device according to Embodiment 6 of the present invention. FIG. 5 is a block diagram showing a detailed configuration of the speech encoding unit 14 of FIG. In FIG. 1 described in the first embodiment, the determination result of the echo determination unit 12 is directly input to the voice encoding unit 13. However, the echo processing device according to the sixth embodiment is a wireless communication terminal such as a mobile phone. DTX (Discintin) commonly used in audio processing in
The configuration is via a uuos transmission (control) unit 15. This DTX control means 15 turns off the wireless transmission output when the transmission signal is in a silent section, and reduces power consumption in that section. DTX
The transmission control method and the comfort noise by the control means 15 include, for example,
3G TS which is a standard for third generation mobile phones
093 "AMR Speech Codec; Sour
ce Controlled Rate Operat
ion ".

The operation of the sixth embodiment according to the present invention will be described below with reference to FIGS. 4 and 5, the same reference numerals as those shown in FIGS. 1 and 2 denote the same or corresponding parts, and a description of the same operations will be omitted. In FIG. 4, reference numeral 14 denotes voice coding means for inputting control information output from the DTX control means 15, 16 denotes modulation means for inputting the coded data CS and flag information DTX and modulates them into a radio signal, and 17 denotes a radio signal. And a demodulation means for outputting encoded data CR. In FIG. 5,
Reference numeral 62 denotes a background noise generation unit that inputs the flag information DTX from the DTX control unit 15. Reference numeral 20 denotes demodulation means for demodulating coded data and flag information DTX from a radio signal on the far end side, and reference numeral 21 denotes modulation means for modulating coded data on the far end side and wirelessly outputting the same.

The DTX control means 15 is a sound / silence determination means 4
And the determination result of the echo determination means 12 is input. If the determination result of the echo determination means 12 indicates an echo residual frame, or if the determination result of the voiced / silent determination means 4 is silent, the flag information DTX is transmitted wirelessly. It is set to “0” meaning transmission off and output to the voice encoding means 14 and to the modulation means 16. The DTX control means 15
If the result of the determination by the echo determination means 12 is not an echo residual frame and the result of the determination by the voiced / silence determination means 4 is voiced, the flag information DTX is set to "1" which means that wireless transmission is on, and the Output to the encoding means 14 and output to the modulation means 16.

The voice coding means 14 generates a residual signal SA (i)
In accordance with the flag information DTX from the DTX control unit 15, the voiced / silent determination result from the voiced / silent determination unit 4, and the echo determination result from the echo determination unit 12, It has a function to generate data. Hereinafter, the operation of the speech encoding unit 14 will be described in detail with reference to FIG.

When the flag information DTX is "1" (wireless transmission ON), the voice coding means 14 codes the received signal SA (i) by the CELP method described in the first embodiment and modulates the coded data CS. Output to means 16. The modulating means 16 continuously modulates the encoded data and outputs it wirelessly.

The speech encoding means 14 outputs the flag information D
When TX is “0” and the echo determination result indicates that the frame is not an echo residual frame, a DTX processing mode is entered, for example, 3G TS 26.092 “AMR Sp
ech Codec; Comfort noise
Aspects ", the background noise is intermittently coded by a method as shown in FIG.
Outputs the encoded LSP input from the linear prediction analysis means 56 to the multiplexing means 61, for example, every six frames.
Further, the power of the residual signal SA (i) is obtained and encoded for each of the same six frames, and output to the multiplexing means 61. The multiplexing means 61 multiplexes the coded data and outputs the multiplexed data to the modulation means 16, and the modulation means 16 intermittently (every six frames) modulates the coded data and wirelessly outputs the modulated data. The modulating means always modulates the flag information DTX and outputs it wirelessly. This saves power consumption in the modulating means 16.

The radio signal modulated by the modulating means 16 is input to the far-end demodulating means 20 via the communication path, and is demodulated into coded data and flag information DTX by the demodulating means 20. Is input to At this time, the encoded data is intermittently input to the audio decoding unit 18. The audio decoding means 18 enters the DTX processing mode when the input flag information DTX is “0”, and for example, 3G TS 26.0992 “AMR Spec”
h Codec; Comfort noiseaspe
cts ", a signal corresponding to background noise is decoded using the received LSP and power.

The voice encoding means 14 outputs the flag information D
If TX is “0” and the echo determination result indicates an echo residual frame, the DTX processing mode involving background noise generation is entered. In this DTX processing mode, as shown in FIG. 6, the background noise generator 62 randomly selects a plurality of LSPs and powers from a set of LSPs and powers previously obtained and stored in a silent section by random numbers using a random number. 6
After averaging in 3, the data is encoded by the encoding means 64 and output to the multiplexing means 61 every six frames. At this time, for example, if N sets of LSPs and powers are stored in total, a number M smaller than N is selected at random. The multiplexing means 61 multiplexes the encoded data and modulates the
The modulation means 16 intermittently (every six frames) modulates this encoded data and outputs it wirelessly.

The background noise generation means 62 outputs the LS input from the linear prediction analysis means 56 when the sound / non-speech judgment result is silence regardless of the value of the flag information DTX and the echo judgment result.
P is always stored in advance for the last few frames.
Signal SA in the same analysis frame storing
The power of (i) is obtained and stored.

The speech decoding means 18 intermittently inputs the encoded data via the communication path and the demodulation means 20.
In the processing mode, a signal corresponding to background noise is decoded using the received LSP and power. Then L
Since the SP and the power are randomly selected and averaged from those previously obtained in the analysis frame of the silent section, the decoded signal decoded by the voice decoding means 18 on the far end side has a moderately changing background. It has spectral characteristics as noise and has good quality.

As described above, according to the sixth embodiment, when the echo determination result is an echo residual frame, the speech encoding means 14 uses the DTX control method generally used for speech encoding for wireless communication. Is used to randomly select and sequentially output the power of the LSP and the received signal obtained in the silent section in advance to generate encoded data corresponding to the background noise. By using a simple configuration and method without providing any simple means, it is possible to generate high-quality pseudo background noise in a section in which only an echo exists in a transmission signal.

Embodiment 7 The echo processing apparatus according to the sixth embodiment randomly selects and averages the LSP and the power obtained in advance in the analysis frame of the silent section every six frames in the echo residual frame. However, a new LSP may be generated every six frames by slightly fluctuating the value of each dimension of the LSP obtained in the first analysis frame determined as the echo residual frame using random numbers.

As described above, according to the seventh embodiment, the intermittently transmitted LSP is made to change slightly, so that there is no large fluctuation in the section where only the echo exists in the transmission signal. High quality pseudo background noise can be generated.

[0064]

As described above, according to the present invention, there is provided an echo judgment means for judging that only a near-end background noise component exists in a transmission signal and an echo signal in which a reception signal is echoed by an echo path. In the analysis frame in which the transmission signal is encoded and only the near-end background noise component and the echo signal are present in the transmission signal by the echo determination means, the code determined in advance in the analysis frame determined as the noise section is used. Is configured to include a speech encoding unit that generates and outputs encoded data corresponding to background noise using analysis parameters for quantization, so that there is no need to provide a new unit for generating pseudo background noise, and a simple With the configuration, there is an effect that high-quality pseudo background noise can be generated in a section in which only an echo exists in a transmission signal.

According to the present invention, the speech encoding method is CE
It is an LP method, and the encoded data corresponding to the background noise is configured to include the speech parameter using the spectrum parameter, the adaptive excitation code, the driving excitation code, and the gain of these two excitation codes. There is an effect that it can be easily applied to the CELP system which is widely used as a voice codec system.

According to the present invention, in the analysis frame in which only the near-end background noise component and the echo signal are present in the transmission signal by the echo determination means, a plurality of analysis frames previously determined to be noise sections are obtained. And a speech encoding unit for randomly selecting a plurality of encoded spectral parameters and sequentially outputting the selected encoded spectral parameters. There is an effect that it is possible to generate a high-quality pseudo background noise having characteristics as a background noise without any noise.

According to the present invention, in the analysis frame in which the echo determination means determines that only the near-end background noise component and the echo signal are present in the transmission signal, the analysis frame is obtained in advance from a plurality of analysis frames that are determined to be noise sections. Speech coding means that randomly selects and encodes a plurality of spectral parameters and sequentially outputs the plurality of spectral parameters, so that the spectral parameters obtained in different analysis frames and the relevance with other parameters are used to encode the spectral parameters. There is also an effect that can be applied to the case where a method of performing the above is used.

According to the present invention, in the analysis frame for which it is determined that only the near-end background noise component and the echo signal are present in the transmission signal by the echo determination means, the spectrum of a plurality of analysis frames previously determined to be noise sections is determined. Since the apparatus is provided with an audio encoding means for averaging the parameters, fluctuating and encoding the values of each dimension of the averaged spectral parameters, and sequentially outputting them, a high quality pseudo background noise without large fluctuation is generated. There is an effect that can be.

According to the present invention, the adaptive excitation code, the driving excitation code, and the coded gain corresponding to both excitation codes, which are determined in the same analysis frame in which the spectral parameters randomly selected are determined, are selected. In addition, since the apparatus is configured to include the audio encoding means for sequentially outputting, it is possible to generate high quality pseudo background noise having the spectral characteristics and power of the actual background noise.

According to the present invention, in the analysis frame in which only the near-end background noise component and the echo signal are present in the transmission signal in the echo determination means, a plurality of analysis frames previously determined to be noise sections are used. A plurality of determined adaptive excitation codes and a set of encoded gains corresponding to the respective adaptive excitation codes are randomly selected and sequentially output,
Further, there is provided a speech encoding unit which randomly selects a plurality of drive excitation codes obtained in a plurality of analysis frames determined in advance as noise sections and sets of encoded gains corresponding to the respective drive excitation codes, and sequentially outputs the selected sets. With this configuration, there is an effect that it is possible to generate high-quality pseudo-background noise with large randomness without biased characteristics.

According to the present invention, the gain of the adaptive excitation code and the gain of the driving excitation code are obtained and retained before encoding, and are provided with a speech encoding means for encoding and outputting each time they are selected. With this configuration, there is an effect that it is possible to cope with a case where a method of encoding a gain using the gain obtained in different analysis frames and the relationship with other parameters is used.

According to the present invention, the gains of the selected adaptive excitation code and driving excitation code are corrected based on the average power of the target signal obtained in advance in a plurality of analysis frames determined as noise sections, and the coding is performed. Is configured to include the voice encoding means for outputting the pseudo-background noise with no large power fluctuation.

According to the present invention, when the gain of the driving excitation code is obtained in the analysis frame previously determined to be a noise section,
Since the adaptive excitation code gain is configured to include a speech encoding unit that obtains the gain of the driving excitation code after substituting a value of zero to a value close to zero, a pseudo background of good quality regardless of the contents of the adaptive excitation codebook. There is an effect that noise can be generated.

According to the present invention, in the analysis frame in which it is determined that only the near-end background noise component and the echo signal are present in the transmission signal by the echo determination means, the DTX silence flag is output and the DTX silence flag is output. Since it is configured to include a voice encoding unit that operates internally in the section processing mode, a simple configuration without using a new unit for generating pseudo background noise using a commonly used DTX control method is provided. The method has an effect that high-quality pseudo background noise can be generated.

According to the present invention, there is provided a speech encoding means for randomly selecting, averaging and encoding a plurality of spectrum parameters obtained in advance in a plurality of analysis frames determined as noise sections, and sequentially outputting the speech parameters. Therefore, there is an effect that it is possible to generate high-quality pseudo background noise having a spectral characteristic as background noise that changes appropriately.

According to the present invention, when the transmission signal contains a large amount of echo components based on the determination result of the echo determination means, the encoding means replaces the encoded data of the transmission signal with the background data. Since the apparatus is provided with background noise generation means for outputting coded data corresponding to noise stored in advance as transmission data, even if a transmission signal contains a large amount of echo components, a code corresponding to background noise is generated. There is an effect that the converted data can be output as transmission data.

[Brief description of the drawings]

FIG. 1 is a block diagram of an echo processing device according to Embodiments 1 to 5 of the present invention.

FIG. 2 is a block diagram of a speech encoding unit provided in the echo processing device according to the first to fifth embodiments of the present invention.

FIG. 3 is an operation explanatory diagram illustrating an operation of generating background noise according to Embodiments 1 to 5 of the present invention.

FIG. 4 is a block diagram of an echo processing device according to a sixth embodiment of the present invention.

FIG. 5 is a block diagram of a speech encoding unit provided in an echo processing device according to a sixth embodiment of the present invention.

FIG. 6 is an operation explanatory diagram illustrating an operation of generating background noise according to Embodiment 6 of the present invention;

FIG. 7 is a block diagram of a conventional echo processing device.

[Explanation of symbols]

REFERENCE SIGNS LIST 1 pseudo echo generation means, 2 addition means, 4 voiced / silence determination means, 10 voice decoding means, 11 voice coding / decoding means, 12 echo determination means, 13 voice coding means, 14 voice coding means, 15 DTX Control means, 1
6 modulation means, 17 demodulation means, 51 adaptive excitation codebook, 52 driving excitation codebook, 53 amplifier, 54 amplifier, 55 addition means, 56 linear prediction analysis means, 57
Synthesis filter means, 58 addition means, 59 sound source search means, 60 background noise generation means, 61 multiplexing means, 62
Background noise generating means, 63 averaging means, 64 coding means, 65 selecting means.

 ──────────────────────────────────────────────────続 き Continuing on the front page (72) Inventor Hirohisa Tazaki 2-3-2 Marunouchi, Chiyoda-ku, Tokyo Mitsui Electric Co., Ltd. (72) Inventor Fumihiro Matsuoka 2-3-2 Marunouchi, Chiyoda-ku, Tokyo F term in Mitsubishi Electric Corporation (reference) 5D020 CC06 5D045 CA01 5K027 AA11 AA16 BB03 DD10 DD18 5K046 HH11 HH79

Claims (13)

[Claims] 1. An echo determining means for determining that only a near-end background noise component and an echo signal in which a received signal is echoed by an echo path are present in a transmitted signal, and the transmission signal is encoded, and the echo signal is encoded. In the determination unit, in the analysis frame determined that only the near-end background noise component and the echo signal are present in the transmission signal, an analysis parameter for encoding obtained in advance in the analysis frame determined to be a noise section is used. And an audio encoding means for generating and outputting encoded data corresponding to background noise. 2. The speech encoding method according to claim 1, wherein the speech encoding method of the speech encoding means is CE.
2. The echo processing apparatus according to claim 1, wherein the coded data corresponding to the background noise is an LP coding system, and the coded data corresponding to the background noise is composed of a spectrum parameter, an adaptive excitation code, a driving excitation code, and a gain of both excitation codes. 3. An analysis frame in which only a near-end background noise component and an echo signal are present in the transmission signal by the echo determination means, a plurality of analysis frames previously determined from a plurality of analysis frames determined to be noise sections. 2. The method according to claim 1, wherein the coded spectral parameters are randomly selected and sequentially output.
Or the echo processing apparatus according to claim 2. 4. An analysis frame in which it is determined that only a near-end background noise component and an echo signal are present in a transmission signal by the echo determination means, a plurality of analysis frames previously determined from a plurality of analysis frames determined to be noise sections. 3. The echo processing apparatus according to claim 1, wherein the spectral parameters are randomly selected, encoded, and sequentially output. 5. An analysis frame in which only a near-end background noise component and an echo signal are present in the transmission signal by the echo determination means, averages the spectral parameters in a plurality of analysis frames previously determined to be noise sections. 3. The echo processing apparatus according to claim 1, wherein the dimensional values of the spectral parameters that have been converted and averaged are fluctuated and encoded for each analysis frame and sequentially output. 6. An adaptive excitation code, a driving excitation code, and an encoded gain corresponding to both excitation codes determined in the same analysis frame as the analysis frame from which the selected spectral parameter is determined, and sequentially output. The echo processing device according to any one of claims 2 to 5. 7. An analysis frame in which only a near-end background noise component and an echo signal are present in the transmission signal by the echo determination means, a plurality of analysis frames previously determined from a plurality of analysis frames determined to be noise sections. An adaptive excitation code and a set of coded gains corresponding to the respective adaptive excitation codes are randomly selected and sequentially output, and a plurality of driving excitation codes and a plurality of driving excitation codes determined in advance in a plurality of analysis frames determined to be noise sections in advance. The echo processing apparatus according to any one of claims 2 to 5, wherein a set of coded gains corresponding to the driving excitation code is randomly selected and sequentially output. 8. The echo processing apparatus according to claim 6, wherein a gain of the adaptive excitation code and a gain of the driving excitation code are obtained and held before encoding, and are encoded and output each time they are selected. . 9. A method for modifying the gains of a selected adaptive excitation code and a driving excitation code based on an average power of a target signal obtained in a plurality of analysis frames previously determined to be a noise section, encoding and outputting the modified excitation code and the driving excitation code. The echo processor according to any one of claims 6 to 8. 10. The gain of a driving excitation code after obtaining a gain of an adaptive excitation code from zero to a value close to zero when obtaining the gain of the driving excitation code in an analysis frame previously determined to be a noise section. The echo processing device according to the above. 11. An analysis frame in which it is determined that only a near-end background noise component and an echo signal are present in the transmission signal by the echo determination means, outputs a silence flag in the DTX control means, and outputs a silence flag in the DTX control means. 2. The echo processing apparatus according to claim 1, further comprising a speech encoding unit that operates internally in a section processing mode. 12. The echo processing apparatus according to claim 1, further comprising a voice coding unit for randomly selecting a plurality of spectral parameters obtained in a plurality of analysis frames determined in advance as noise sections, averaging and encoding the resulting parameters, and sequentially outputting the selected parameters. apparatus. 13. An echo determination means for determining a state in which a transmission signal contains more echo components of a received signal reflected by an echo path than voice components of a near-end speaker, and encoding the transmission signal. Encoding means for outputting the encoded signal, based on the determination result of the echo determining means, when the transmission signal contains a large amount of echo components. An echo processing apparatus, comprising: a background noise generating unit that outputs, as transmission data, encoded data corresponding to background noise instead of data as transmission data.