RU2520402C2 - Multi-resolution switched audio encoding/decoding scheme - Google Patents

Abstract

FIELD: physics, acoustics.

SUBSTANCE: invention relates to audio encoding technologies. An audio encoder for encoding an audio signal has a first coding channel for encoding an audio signal using a first coding algorithm. The first coding channel has a first time/frequency converter for converting an input signal into a spectral domain. The audio encoder also has a second coding channel for encoding an audio signal using a second coding algorithm. The first coding algorithm differs from the second coding algorithm. The second coding channel has a domain converter for converting an input signal from an input domain into an output domain audio signal.

EFFECT: improved encoding/decoding of audio signals in low bitrate circuits.

21 cl, 43 dwg, 10 tbl

Description

The present invention relates to audio coding, and especially to low bitrate coding schemes.

Encoding schemes for the frequency domain, such as MP3 or AAS, are known in the art. These encoders in the frequency domain are based on the “time domain region” transformation, a subsequent quantization step in which the quantization error is controlled using information from the perceptual module, and an encoding step in which the quantized spectral coefficients and corresponding third-party information are encoded without loss of information using code tables.

On the other hand, there are encoders that are very well suited for speech processing, such as AMR-WB +, described in 3GPP TS 26.290. Such speech coding schemes perform linear predictive filtering of a signal over a time interval. Linear predictive filtering is obtained from a linear analysis of the prediction of the input signal over a time interval. The resulting linear predictive (LP) filter coefficients are quantized / encoded and transmitted as third-party information. The process is known as linear prediction coding (LPC). A prediction difference signal or a prediction error signal is generated at the filter output, which is also known as an excitation signal encoded using the steps of the synthesis analysis of the ACELP encoder or, alternatively, encoded using an encoder based on the Fourier transform with an overlay. The decision between ACELP coding and transformed drive signal coding, also called TLC coding, is made using a closed open loop algorithm.

Frequency-domain audio coding schemes, such as a high-performance AAC (non-ACC) coding scheme that combines AAC coding and spectral band repetition (SBR) techniques, and can be combined with a combined stereo or multi-channel coding instrument, which is known as "MPEG surround "

On the other hand, speech encoders such as AMR-WB + also have a high-frequency extension level and stereo functionality.

Frequency domain coding schemes show high quality at low bitrate for music signals. The problem, however, is the quality of speech signals at low bitrates.

Speech coding schemes show high quality for speech signals even at low bitrate, but show low quality for other signals at low bitrate.

An object of the present invention is to provide a concept for improved coding / decoding.

This is achieved by an audio encoder in accordance with claim 1, an audio encoding method in accordance with claim 9, a decoder in accordance with claim 10, a decoding method in accordance with claim 19, an encoded message in accordance with claim 20, or a computer program in accordance with paragraph 21 of the claims.

The present invention is based on the idea that a hybrid or dual-mode switchable coding / decoding scheme is advantageous since a better coding algorithm may be selected for a particular signal feature. In other words, the present invention does not seek a signal coding algorithm that is perfectly matched to all signal features. Such a scheme would always be a compromise, which can be seen from the enormous technical differences between the properties of audio encoders, on the one hand, and speech encoders, on the other. Instead, the present invention combines various encoding algorithms, such as an algorithm for encoding speech, on the one hand, and an algorithm for encoding audio, on the other hand, within a switching circuit so that for each part of the audio signal, an appropriate encoding algorithm is optimally selected. In addition, it is also a feature of the present invention that both coding channels include a time / frequency converter, but a further signal converter, such an LPC processor, is provided in one coding channel. This converter serves to confirm that the second coding channel is better suited for a particular signal feature than the first coding channel. However, this is also a feature of the present invention, the signal at the output of the processor is also converted to a spectral representation.

Both transducers, that is, the first transducer in the first coding channel and the second transducer in the second coding channel, are configured to perform multi-resolution transform coding, where the resolution of the corresponding transducer is set depending on the audio signal, and in particular, depending on the audio signal, in fact encoded in the corresponding coding channel so as to achieve a good compromise between quality, on the one hand, and bitrate, on the other hand, or Points of view of a specific constant quality, to achieve the lowest bit rate, or in terms of a constant bit rate to get the best quality.

In accordance with this invention, the time / frequency resolution of these two converters can be set preferably independently of each other so that each time the converter can be optimally matched to the time / frequency resolution requirements of the corresponding signal. The efficiency of bits, that is, the ratio between useful bits, on the one hand, and bits of third-party information, on the other hand, is higher for large block sizes / window lengths. Therefore, it is preferable that both transducers operate over a longer window, since basically the same amount of third-party information relates to the long time portion of the audio signal and the shorter block sizes / window lengths. It is desirable that the resolution of time / frequency in the encoding channels could also be influenced by other encoding / decoding tools located in these channels. It is desirable that the second coding channel including a signal converter, such as an LPC processor, include another hybrid circuit, such as an ACELP channel, on the one hand, and a TLC circuit, on the other hand, where the second converter is included in a TLC circuit. It is desirable that the resolution of the time / frequency converter located in the TLC channel is also influenced by the encoding decision, so that part of the signal in the second encoding channel is processed in the TLC channel having a second converter or in the ACELP channel without a time / converter frequency.

Basically, neither the signal converter, nor the second coding channel, and in particular, the first processing channel in the second coding channel and the second processing channel in the second coding channel, should not be speech-related elements, such as an LPC analyzer for the signal converter, the encoding device TLC for the second processing channel and the ACELP encoder for the first processing channel. Other solutions are also useful when other features of an audio signal other than speech, on the one hand, and music, on the other hand, are evaluated. Any signal converters and coding channels can be used, and the best suitable algorithm can be found by the synthesis analysis circuit so that, on the encoder side, all coding alternatives are drawn for each part of the audio signal and the best result is selected where the best result can be found using the objective function to the coding results. Then, the identification of the third-party decoder information underlying the encoding algorithm for a specific part of the encoded audio signal is attached to the encoded audio signal by the output interface encoder so that the decoder does not care about any decisions about the encoder side or any signal features, but simply selects coding channel, depending on the transmitted third-party information. In addition, the decoder will not only select the correct decoding channel, but will also select, based on the third-party information encoded in the encoded message, which time / frequency resolution should be applied in the corresponding first decoding channel and the corresponding second decoding channel.

Thus, the present invention provides an encoding / decoding scheme that combines the advantages of various encoding algorithms and avoids the disadvantages of these encoding algorithms that occur when a portion of a signal would have to be encoded by an algorithm that does not correspond to the current encoding algorithm. In addition, this invention avoids any inconvenience that occurs if there are different requirements for the resolution of time / frequency when processing different parts of the audio signal in different coding channels. Instead, due to the variable time / frequency resolution of the time / frequency converters in both channels, any distortion that would occur in a scenario where the same time / frequency resolution is used for both coding channels, or in which only a constant time / frequency resolution is possible for any coding channels are at least reduced or even completely eliminated.

The second switch again switches between the two processing channels, but in a region different from the "external" region of the first channel. Again, the operation of one “internal” channel is mainly determined by the original model or SNR calculations, and the other “internal” channel can be determined by the model of the hearing (auricle) and / or the psychoacoustic model, that is, masking or at least incorporating coding aspects in the frequency / spectral region. For example, one “internal” channel has a frequency domain converter / spectral converter, and the other channel has a device encoding in another region, such as an LPC region, and this is an encoding device, for example, CELP or ACELP with processing of the input signal without spectral transformations.

A further preferred embodiment is an audio encoder comprising a first information channel, such as spectral domain encoding, a second information source or SNR oriented encoding channel, such as an LPC region encoding channel, and a switch in order to switch between the first encoding channel and a second coding channel, the second coding channel including a converter in a region other than a time interval, such as an LPC analysis step generating a signal and where the second coding channel also includes a special area, such as an LPC region processing channel, and a special spectral region, such as an LPC spectral processing channel, and an additional switch to switch between a special coding channel and a special spectral channel processing.

A further embodiment of the invention is an audio decoding apparatus including a first region, such as a spectral decoding channel, a second region, such as an LPC decoding channel, in order to decode a signal, such as an excitation signal in a second region, and a third region, such as LPC- a spectral channel of a decoder, in order to decode a signal, such as an excitation signal in a third region, such as an LPC spectral region, where the third region is obtained by performing frequency conversion from the second region, where a first switch for a second region signal and a third region signal is provided, and where a second switch is provided for switching between a decoder for the first region and a decoder for the second region or third region.

Preferred solutions of the present invention are described in the attached drawings, where:

Figa is a block diagram of an encoding device in accordance with a first aspect of the present invention;

Fig. 1b is a block diagram of a decoding apparatus in accordance with a first aspect of the present invention;

Fig. 1c is a block diagram of an encoding device in accordance with a further aspect of the present invention;

Fig. 2a is a block diagram of an encoding device in accordance with a second aspect of the present invention;

Fig.2b is a schematic diagram of a decoding device in accordance with a second aspect of the present invention;

Fig. 2c is a block diagram of an encoding device in accordance with a further aspect of the present invention;

Fig. 3a illustrates a block diagram of an encoding apparatus in accordance with a further aspect of the present invention;

Fig. 3b illustrates a block diagram of a decoding apparatus in accordance with a further aspect of the present invention;

Fig. 3c illustrates a schematic representation of an encoding device / method with cascading switches;

Fig. 3d illustrates a schematic diagram of a device or decoding method in which combiner stages are used;

Fig. 3e illustrates a signal in a time slot and a corresponding representation of an encoded signal illustrating short intersecting regions that are included in both encoded signals;

Fig. 4a illustrates a block diagram with a switch placed in front of coding channels;

Fig. 4b illustrates a block diagram of an encoding device with a switch located behind the encoding channels;

Fig. 5a illustrates the waveform of a speech segment over a time interval as a quasiperiodic or pulse-like signal segment;

Fig. 5b illustrates the spectrum of the segment of Fig. 5a;

Fig. 5c illustrates a non-voice segment in a time interval, such as, for example, a segment similar to noise;

Fig. 5d illustrates the spectrum of the time interval of Fig. 5c;

6 illustrates a block diagram of a CELP synthesis analysis encoder;

Figures 7a and 7d illustrate voice / non-voice excitation signals, such as, for example, pulse-like signals;

Fig. 7e illustrates a portion of an encoder LPC stage providing short-term prediction information and a prediction (excitation) error signal;

Fig.7f illustrates a further embodiment of the LPC device in order to generate a weighted signal;

FIG. 7g illustrates a device for converting a weighted signal to an excitation signal by applying the inverse “weighting” operation and then analyzing the excitation as required in transducer 537 in FIG. 2b;

Fig. 8 illustrates a block diagram of a combined multi-channel algorithm in accordance with a solution of the present invention;

9 illustrates a preferred solution for a bandwidth extension algorithm;

Fig. 10a illustrates a detailed description of a switch performing an open loop solution; and illustrates a switch operating in a closed loop algorithm;

Fig. 11A illustrates a block diagram of an audio encoder in accordance with another aspect of the present invention;

11B illustrates a block diagram of another solution of the proposed audio decoder;

Figa illustrates another solution of the proposed encoding device;

Figv illustrates another solution of the proposed decoder;

13A illustrates the relationship between resolution and window / transform length;

13B illustrates an overview of a series of transform windows for a first coding channel and a transition from a first to a second coding channel;

13C illustrates many different window sequences, including a window sequence for a first coding channel and a sequence for transitioning to a second channel;

Figa illustrates the creation of a preferred solution for the second coding channel;

14B illustrates short windows applied in a second coding channel;

Fig. 14C illustrates medium-sized windows applied in a second coding channel;

Fig.14D illustrates the long windows used in the second coding channel;

Fig. 14E illustrates a typical sequence of ACELP frames and TLC frames within a superframe;

Fig. 14F illustrates various transform lengths corresponding to different time / frequency resolutions for a second encoding channel; and

Fig.14G illustrates the construction of the window using the definitions of fig.14F.

11A illustrates a solution of an audio encoder in order to encode an audio signal. The encoder includes a first encoding channel 400 in order to encode the audio signal using the first encoding algorithm to obtain the first encoded signal.

The audio encoder further includes a second encoding channel 500 in order to encode the audio signal using the second encoding algorithm to obtain a second encoded signal. The first coding algorithm is different from the second coding algorithm. Additionally, the first switch 200, configured to switch between the first coding channel and the second coding channel so that for the portion of the audio signal, either the first encoded signal or the second encoded signal is in the encoded output signal 801.

The audio encoder illustrated in FIG. 11A further includes a 300/525 signal analyzer, which is configured to analyze a portion of the audio signal to determine whether a portion of the audio signal is represented as a first encoded signal or a second encoded signal in encoded output signal 801.

The signal analyzer 300/525 is further configured to determine the corresponding non-constant time / frequency resolution of the first converter 410 in the first coding channel 400 or the second converter 523 in the second coding channel 500. This resolution is applied when the first encoded signal or second encoded signal is generated representing a portion of an audio signal.

The audio encoder further includes an output interface 800 to generate an encoded output signal 801 consisting of an encoded representation of a portion of the audio signal and information indicating whether the representation of the audio signal is a first encoded signal or a second encoded signal and an indication of time / frequency resolution used to decode the first encoded signal and the second encoded signal.

The second encoding channel usually differs from the first encoding channel in that the second encoding channel further includes a region converter in order to convert the audio signal from the region in which the audio signal is processed in the first encoding channel to another region. Typically, the area converter is an LPC 510 processor, but the area converter can be implemented in any other way, as long as the area converter is different from the first converter 410 and the second converter 523.

The first converter 410 is a time / frequency converter, typically including a window driver 410a and a converter 410b. Window driver 410a applies the analytic window to the input audio signal, and converter 410b converts the signal in the generated window into a spectral representation.

Similarly, the second transducer 523 typically includes a window shaper 523a in series with the transducer 523b. Window driver 523a receives the signal generated by converter 510 and generates a signal representation processed by the window function. The output of the window driver 523a is provided to a converter 523b to form a spectral representation. The converter may be an FFT or, preferably, an MDCT processor implementing an appropriate algorithm using software or hardware or mixed use of hardware / software. Alternatively, the converter may be a filter bank configured as a QMF filter bank, which may be based on real or complex modulation of a prototype filter. To implement a specific filter bank, a window is used. However, to implement another filter bank, processing by the window function, as required for the FFT MDCT based conversion algorithm, is not necessary. When a filter bank is used, then the filter bank has a variable resolution, and this resolution controls the frequency resolution of the filter bank, and additionally, the time resolution or only the frequency resolution. However, when the converter is implemented as an FFT or MDCT or any other suitable converter, then the frequency resolution is related to the time resolution, while an increase in the frequency resolution obtained with a large time block size automatically corresponds to a lower time resolution and vice versa.

Additionally, the first encoding channel may include a quantization / encoding unit 421, and the second encoding channel may also include one or more further encoding tools 524.

It is important that the signal analyzer is configured to generate a resolution control signal for the first converter 510 and for the second converter 523. Thus, independent resolution control is implemented in both coding channels to have an encoding circuit that, on the one hand, provides a low bit rate, and on the other hand, it provides maximum quality at a low bit rate. To achieve a lower bitrate, longer windows or longer conversion lengths are required, but in situations where these large lengths lead to distortion due to the low temporal resolution, shorter window lengths and shorter conversion lengths that result in lower frequency resolution are applied. Preferably, the signal analyzer uses statistical analysis or any other analysis that is suitable for the corresponding algorithms in the coding channels. In one embodiment, in which the first coding channel is a coding channel in the frequency domain, such as an AAC-based coding device, and in which the second coding channel includes an area converter in the form of an LPC processor 510, the signal analyzer, by controlling switch 200, performs speech separation / music so that the speech portion of the audio signal enters the second coding channel. The musical portion of the audio signal controlled by the switch 200, as indicated by the control lines, enters the first coding channel 400. Alternatively, as will be discussed later in FIG. 1C or FIG. 4B, the switch may also be placed in front of the output interface 800.

In addition, the signal analyzer can receive an audio signal input to the switch 200, or an audio signal generated by the switch 200. In addition, the signal analyzer performs an analysis to not only provide the audio signal to the corresponding coding channel, but also determine the appropriate time resolution / the frequency of the corresponding converter in the corresponding coding channel, such as the first converter 410 and the second converter 523, as indicated by resolution control lines connecting the signal analyzer and a converter.

11B includes a preferred embodiment of an audio decoder corresponding to the audio encoder of FIG. 11A.

The audio decoder of FIG. 11B is configured to decode an encoded audio signal, such as encoded output signal 801, the generated output interface 800 of FIG. 11A. The encoded signal includes a first encoded audio signal encoded in accordance with a first encoding algorithm, a second encoded signal encoded in accordance with a second encoding algorithm, a second encoding algorithm different from the first encoding algorithm, and information indicating whether a first encoding algorithm or a second encoding algorithm is used an algorithm for decoding a first encoded signal and a second encoded signal, and time / frequency resolution information for the first is encoded a second audio signal and a second encoded audio signal.

The audio decoder includes a first decoding channel 431, 440 in order to decode the first encoded signal based on the first encoding algorithm. In addition, the audio decoder includes a second decoding channel in order to decode the second encoded signal using a second encoding algorithm.

The first decoding channel includes a first controllable converter 440 configured to convert from a spectral region to a time interval. The controlled converter is adapted to be controlled using time / frequency resolution information from a first encoded signal to obtain a first decoded signal.

The second decoding channel includes a second controllable converter configured to convert from a spectral representation to a temporal representation, a second controllable converter 534 configured to control using time / frequency resolution information 991 for the second encoded signal.

The decoder further includes a controller 990 in order to control the first converter 540 and the second converter 534 in accordance with the time / frequency resolution information 991.

In addition, the decoder includes a region converter in order to generate a synthesized signal using a second decoded signal to perform the inverse transform performed by the region converter 510 in the encoder of FIG. 11A.

Typically, the region converter 540 is an LPC synthesizing processor controlled using the LPC filter information included in the encoded message, where this LPC filter information was generated by the LPC processor 510 in FIG. 11A and was input to the encoder output as third-party information. The audio decoder finally includes a combiner 600 to combine the first decoded signal generated by the first transformer of the region 440 and the synthesized signal to obtain a decoded audio signal 609.

In the proposed embodiment, the first decoding channel further includes a decanter / decoder 431 in order to perform conversions that are inverse to those performed by the corresponding block 421 of the encoding device. However, it is clear that quantization cannot be completely reversible, since it is an operation with information loss. However, the dequantizer completely reverses a certain heterogeneity of quantization, such as logarithmic quantization or quantization with multiplexing.

In the second decoding channel, the corresponding step 533 is applied in order to carry out the operations inverse to that applied in step 524. It is preferable that step 524 includes uniform quantization. Therefore, the corresponding stage 533 will not have a specific stage of dequantization in order to remove homogeneous quantization.

The first converter 440, like the second converter 534, may include corresponding inverters 440a, 534a, synthesis windows 440b, 534b connected in series with the overlap / add unit 440c, 534c. Overlapping / adding blocks are required when converters are used, and more specifically, inverters 440a, 534a, use a combination of input transforms, such as a modified discrete cosine transform. Then, the overlap / add operation will cancel time slot combining (TDAC). However, when converters are used that do not use transform combining, such as inverse FFT, an overlap / add block 440s is not required. In this design, a fade overlay or crossfade operation can be applied to avoid distortion caused by blocking.

Similarly, combiner 600 can be a switchable combiner, or provide a crossfade, or use combining to avoid blocking distortions when the combiner transitions to the window function processing, similar to a block overlapping / adding within the current coding channel.

Figa illustrates a solution to the invention having two stages of switches. A mono signal, a stereo signal, or a multi-channel signal are supplied to switch 200. The switch 300 is controlled by block 300. The input to the selection block is fed to the input of switch 200. Alternatively, decision block 300 may also receive third-party information that is included in the mono signal, stereo signal, or a multichannel signal or, at least, is associated with a signal where this information exists, which can, for example, be generated initially when a mono signal, a stereo signal or a multichannel signal is generated.

The selector / signal analyzer 300 activates a switch 200 to generate a signal in the coding channel of the frequency domain 400, illustrated in the upper part of FIG. 1a, or in the LPC coding channel 500, illustrated in the lower part of FIG. 1a. The main element of the frequency domain coding channel is the spectral transform unit 410, which serves to convert the overall output of the preprocessing stage (as will be discussed later) into the spectral region. The spectral transform block may include an MDCT, QMF algorithm, an FFT algorithm, a wavelet analysis, or a filter bank, such as a filter bank with a critical sampling having a certain number of channels, where the subband signals in this filter bank can be real signals or complex signals. The output of the spectral transform unit 410 is encoded using a spectral audio encoder 421, which may include processing units known from the AAC encoding scheme.

In general, the processing in channel 400 is processing based on a perception model or an information model of hearing. Thus, this channel simulates the human audience sound received by the system. The inverse of this is the processing in channel 500, which is supposed to generate an excitation signal, a differential signal or an LPC region. In general, the processing in channel 500 is processing based on a speech model or an information generation model. For speech signals, this model is the model of a system that forms human speech / sound. If, however, the sound comes from various sources that require different models for generating sound to be encoded, then the processing in channel 500 may be different.

In coding channel 500, the main element is the LPC 510, which generates LPC information that is used to control the parameters of the LPC filter. This LPC information is transmitted to the decoder. The LPC output of processor 510 is an LPC region signal that consists of an excitation signal and / or a weighted signal.

The LPC processor generally generates an LPC region signal, which can be any signal in the LPC region, such as the excitation signal in FIG. 7e, or a weighted signal in FIG. 7f, or any other signal that was generated using the filter LPC coefficients for the audio signal. In addition, the LPC device can also determine these coefficients and can also quantize / code these coefficients.

The decision in the selection unit may be an adaptive signal so that the selection unit performs music / speech separation and controls the switch 200 in such a way that the music signals enter the first channel 400 and the speech signals enter the second channel 500. In one solution, the selection information of the selection block is supplied to the output bit so that the decoder can use this selection information to perform the correct decoding operations.

Such a decoder is illustrated in FIG. 1b. The signal generated by the spectral audio encoder 421 is, after being transmitted, input to the spectral audio decoder 431. The output of the spectral audio decoder 431 is supplied to the converter in the time domain 440. Similarly, the output signal of the LPC encoding channel of region 500 in FIG. 1a is supplied to the decoder and processed by elements 531, 533, 534, and 532 in order to obtain an LPC drive signal. The LPC excitation signal is fed to the synthesis LPC block 540, which receives, at another input, LPC information generated by the corresponding LPC analysis step 510. The output signal is converted to the time domain 440 and / or the output signal from the LPC synthesis block 540 is sent to the switch 600. With the switch 600 controls the control signal of the switch, which was, for example, generated by the block selection / analysis of the signal 300, or which is provided externally by the driver of the original mono signal, a stereo signal or a multi-channel signal. The output of switch 600 is a full mono signal, a stereo signal, or a multi-channel signal.

The input signal of the switch 200 and the signal selection / analysis unit 300 may be a mono signal, a stereo signal, a multi-channel signal, or in general an audio signal. Depending on the choice that can be obtained from the input signal selector 200 or from any external source, such as an original audio signal generator underlying the input signal of the switch 200, a switch is made between the frequency coding channel 400 and the LPC coding channel 500. Channel frequency coding 400 includes a spectral transform unit 410 connected to quantization / coding unit 421. The quantization / encoding unit may include any of the functionalities known from modern frequency range encoders, such as an AAC encoder. In addition, the quantization operation in the quantization / coding unit 421 can be controlled through a physical-acoustic module that generates physical-acoustic information, such as a masking physical-acoustic frequency threshold, which is supplied to block 421.

In the LPC coding channel, the output of the switch is processed by the LPC processor 510, forming third-party LPC information and the signal of the LPC region. The excitation encoder intellectually includes an additional switch in order to switch the further processing of the signal of the LPC region between the quantization / coding operation 522 in the LPC region or the quantization / coding unit 524 that processes the data in the LPC spectral region. To this end, a spectral converter 523 is installed at the input of a quantization / coding unit 524. The switch 521 controls the open loop mode or closed loop mode depending on certain settings, such as, for example, described in the technical specification AMR-WB +.

For the closed loop control mode, the encoder further includes an inverse quantization / encoding transform 531 for the LPC domain signal, an inverse quantization / encoding transform 533 for the LPC domain spectral signal and an inverse spectral converter 534 for the output signal of block 533. The encoded and decoded signals in the second channels processing is provided to switch control device 525. In switch control device 525, these two output signals are compared to each other with a friend and / or with an objective function or with an objective function calculated by comparing the distortion in both signals so that a signal having a lower distortion is used in order to decide how to control switch 521. Alternatively, if both The channel provides inconsistent bitrates, a channel can be selected that provides a lower bitrate, even when the signal-to-noise ratio of this channel is lower than the signal-to-noise ratio of the other channel. Alternatively, the objective function may use the input signal-to-noise ratio of each signal and the bit rate of each signal and / or additional criteria to find the best solution for a specific purpose. If, for example, the goal is such that the bit rate should be as low as possible, then the objective function would be heavily based on the bitrate of the two signals generated by elements 531, 534. However, when the main goal is to have the highest quality for a certain bitrate, then control of the switch 525 could, for example, exclude a signal that has a bitrate higher than the allowable bitrate, and if both signals have a bitrate below the allowable bitrate, the control of the switch would select a signal that has the best wearing signal / noise ratio, that is, having a smaller distortion of the quantization / coding.

The decoding scheme in accordance with this invention, as stated previously, is illustrated in fig.1b. For each of the three possible types of output signal, there is a specific decoding / dequantization unit 431, 531 or 533. While block 431 generates a spectrum of a time interval that is converted to a time interval using a frequency / time converter 440, block 531 generates an LPC- signal areas, and block 533 forms an LPC spectrum. To ensure that the input signals supplied to the switch 532 are in the LPC region, an LPC spectrum / LPC converter 534 is installed. The output of the switch 532 is converted back to the time interval using the LPC synthesis unit 540, which is controlled by the information generated and transmitted by the LPC encoder. Then, behind block 540, there is time slot information in both channels that is switched in accordance with the control signal of the switch to obtain the final audio signal, such as a mono signal, a stereo signal, or a multi-channel signal, which depends on the input signal to the encoding circuit on figa.

Fig. 1c illustrates a further embodiment with a different arrangement of switch 521, similar to the principle illustrated in Fig. 4b.

Fig. 2a illustrates a preferred coding scheme in accordance with a second aspect of the invention. The general pre-processing circuitry connected to the input of the switch 200 may include an surround / combined stereo unit 101, which generates output parameters of the combined stereo and mono output signal, which is generated by down-mixing the input signal, which is a signal having two or more channels. In general, the signal generated at the output of block 101 may also be a signal having more channels, but due to the downmix functionality of block 101, the number of channels at the output of block 101 will be less than the number of input channels to block 101.

The general preprocessing scheme may include, alternatively to block 101, or in addition to block 101, a bandwidth extension block 102. In the solution of FIG. 2a, the output signal of block 101 is fed to a frequency band extension block 102, which, in the encoder of FIG. 2a, generates a limited-band signal, such as a low-frequency signal, at the output. Typically, this signal is sampled at a lower frequency (for example, at a frequency two times lower). In addition, for a high-frequency input signal to block 102, frequency extension parameters, such as spectral envelope parameters, inverse filter parameters, noise level parameters, etc., known from the non-AAC MPEG characteristics, are generated and formatted into the bitstream by multiplexer 800 -four.

Typically, the signal selection / analysis unit 300 receives an input signal to block 101 or block 102 to select between, for example, music mode or speech mode. In music mode, the upper coding channel 400 is selected, while in speech mode, the lower coding channel 500 is selected. Typically, the signal selection / analysis unit further controls the surround / combined stereo unit 101 and / or the band extension block 102 to adapt the functionality of these blocks to a specific signal. Thus, when the signal selection / analysis unit decides that a certain time portion of the input signal relates to the first mode, such as the music mode, the signal selection / analysis unit 300 can control certain features of the block 101 and / or block 102. Alternatively, when the block selecting / analyzing the signal 300 decides that the signal belongs to the speech mode or, generally, to the LPC region, then certain features of the blocks 101 and 102 can be controlled in accordance with the output signal of the signal selection / analysis block.

Preferably, the spectral conversion of the coding channel 400 is done using the MDCT operation, which is even more preferably an MDCT operation with time distortion, where the distortion can be controlled between zero and a high amount of distortion. In zero distortion, the MDCT time distortion operation in block 411 is a conventional MDCT operation known in the art. The amount of time distortion together with third-party information in the distorted time can be transmitted / entered into the bitstream by the multiplexer 800 as third-party information.

In the LPC coding channel, the LPC area encoder may include an ACELP 526 core that calculates pitch transmission, pitch interval and / or codebook information such as codebook index and transmission. TLC mode, known from 3GPP TS 26.290, includes processing a perceptually weighted signal in the transform domain. The Fourier transformed, weighted signal is quantized using a multi-level quantizing lattice (algebraic VQ) and noise quantization factor. The conversion is calculated in windows with a length of 1024, 512, or 256 samples. The excitation signal is reconstructed using reverse filtering by passing a quantized weighted signal through a reverse weighting filter.

In the first coding channel 400, the spectral converter preferably includes an adapted MDCT operation including certain window functions following the quantization / encoding step, which may consist of a single vector quantization step, but is preferably a combined scalar quantization / entropy encoder similar to a quantization / encoding unit in coding channel of the frequency domain, that is, in block 421 in figa.

In the second coding channel, there is an LPC unit 510 followed by a switch 521, followed by an ACELP unit 526 or a TCX unit 527. ACELP is described in 3GPP, TS 26.190 and TLC is described in 3GPP TS 26.290. In general, ACELP unit 526 receives an LPC drive signal, which is calculated by the procedure described in FIG. 7e. TLC block 527 receives a weighted signal, as shown in FIG. 7f.

In TLC, the conversion is applied to a weighted signal calculated by passing an input signal through an LPC-based weighting filter. The weighting filter uses the solution of the invention given by the expression (1-A (z / γ)) / (1-µz ^-1 ). Thus, the weighted signal is the signal of the LPC region, and its conversion is the LPC spectral region. The signal processed by ACELP 526 is an excitation signal and is different from the signal processed by block 527, but both signals are in the LPC region.

, чтобы преобразовать взвешенную область в область возбуждения.On the side of the decoder illustrated in FIG. 2b, after inverse spectral conversion at block 537, the inverse filtering of the weights is applied, so that (1-μz ^-1 ) / (1-A (z / γ)). The signal is then filtered by (1-A (z)) to enter the LPC field of excitation. Thus, the conversion in block 534 of the LPC region and block 537 TLC ^-1 , including the inverse transformation and then filtering by $$\frac{\left(\mathrm{one}-\mu z^{-\mathrm{one}}\right)}{\left(\mathrm{one}A\left(z/\gamma \right)\right)}\left(\mathrm{one}-A\left(z\right)\right)$$

to convert the weighted region to the excitation region.

Although block 510 in FIGS. 1a, 1c, 2a, 2c illustrates a single block, block 510 may generate various signals while these signals are in the LPC area. The actual mode of block 510, such as the drive signal mode or the weighted signal mode, may depend on the actual position of the switch. Alternatively, block 510 may have two parallel processing devices, where one device is implemented similarly to that shown in FIG. 7e and the other device is implemented as shown in FIG. 7f. Therefore, the LPC area at the output of block 510 may represent either an LPC drive signal, or an LPC weighted signal, or any other signal of the LPC area.

Preferably, in the second coding channel (ACELP / TCX) of FIGS. 2a or 2c, the signal is pre-processed by the 1-0.68z ^-1 filter, which generates biases, before encoding. In the ACELP / TCX decoder shown in FIG. 2b, the synthesized signal is processed by an inverse filter 1 / (1-0.68z ~ ¹ ), which eliminates these biases. Predictions can be generated at LPC block 510, where the signal is pre-biased before LPC analysis and quantization. Similarly, the removal of bias can be part of the LPC ^-1 540 block synthesis LPC.

Fig. 2c illustrates a further solution for implementing the device of Fig. 2a, but with a different arrangement of the switch 521, similar to the circuit in Fig. 4b.

In a preferred solution, the first switch 200 (see FIGS. 1a or 2a) is controlled by an open loop solution (as shown in FIG. 4a), and the second switch is controlled by a closed loop solution (as shown in Figure 4b).

For example, in FIG. 2c, a second switch is located after the ACELP and TLC units, as shown in FIG. 4b. Then, in the first processing channel, the first LPC region represents the LPC excitation, and in the second processing channel, the second LPC region represents the LPC weighted signal. Thus, the first signal of the LPC region is obtained by filtering (1-A (z)) to convert it to a difference signal of the LPC region, while the second signal of the LPC region is obtained using the filter (1-A (z / γ)) / (1-µz ^-1 ) to convert the signal to a weighted one in the LPC area.

Fig. 2b illustrates a decoding scheme corresponding to the coding scheme in Fig. 2a. The bitstream generated by the bitstream multiplexer 800 of FIG. 2a is the input bit stream of the demultiplexer 900. Depending on the information received, for example, from the bit stream in the mode determination block 601, the switch is controlled by the decoder 600 to either send signals from the upper channel or signals from the lower channel to block 701 bandwidth extension. The bandwidth extension unit 701 receives third-party information from the bitstream multiplexer 900 and, based on this third-party information and the output signal of the mode determination unit 601, restores the high-frequency band based on the output signal of the low-frequency band of the switch 600.

The full-band signal generated by block 701 is fed to the combined stereo / surround processing unit 702, which restores two stereo channels or several multi-channels. In general, block 702 generates more channels than was entered into this block. Depending on the application, the input to block 702 may even include two channels, such as in stereo mode, and may even include more channels, while there are more channels at the output of this block than at the input to this block.

The switch 200, as shown, is switched between both channels so that only one channel receives the processed signal and the other channel does not receive the processed signal. In an alternative solution, however, the switch may also be installed, for example, behind the audio encoder 421 and the excitation encoder 522, 523, 524, which means that both channels 400, 500 process the same signal in parallel. In order not to double the bitrate, only the signal generated by one of the coding channels 400 or 500 is selected for recording in the output bitstream. The selector will then work so that the signal recorded in the output bitstream minimizes a certain cost function, where the cost function can be generated by the bitrate, or created by perceptual distortion, or by a combined cost function bitrate / perceptual distortion. Therefore, either in this mode or in the mode illustrated in the figures, the selector can also work in a closed-loop method to make sure that only the output signal of the coding channel that has the lowest bitrate for a given perceptual distortion is written to, or, for a given bitrate, has the lowest perceptual distortion. In the closed loop method, the feedback input can be obtained from the output signals of the three quantizer / counter blocks 421, 522, and 424 shown in Fig. 1a.

In a solution having two switches, that is, a first switch 200 and a second switch 521, it is preferable that the time resolution for the first switch is lower than the time resolution for the second switch. Stated differently, the blocks of the input signal to the first switch, which can be switched through a switching operation, are larger than the blocks switched by the second switch operating in the LPC region. For example, a frequency domain / LPC area switched 200 may switch blocks of 1024 samples length, and a second switch 521 may switch blocks having 256 samples each.

Although some of figa-10b are illustrated as block diagrams of the device, these sides are simultaneously an illustration of a method where the functionality of the block corresponds to the steps of the method.

Fig. 3a illustrates an audio encoder generating an encoded audio signal as an output signal of a first encoding channel 400 and a second encoding channel 500. In addition, the encoded audio signal preferably includes third-party information such as preprocessing parameters from a common preprocessing layer or, as discussed in connection with the previous figures, switch control information.

Preferably, the first coding channel serves to encode the audio intermediate signal 195 in accordance with the first coding algorithm, wherein the first coding algorithm has an information model of hearing. The first coding channel 400 generates a first output signal from the encoder, which is an encoded spectral information representation of the intermediate audio signal 195.

In addition, the second encoding channel 500 is configured to encode the intermediate audio signal 195 in accordance with the second encoding algorithm, the second encoding algorithm based on the source information model and generation, in the second output signal of the encoder, the encoded parameters of the source information model represent the intermediate audio signal .

The audio encoder further includes a general preprocessing step for preprocessing the signal 99 to form an intermediate audio signal 195. Specifically, a general preprocessing step serves to process the water signal 99 so that the intermediate audio signal 195, i.e. The result of the general preprocessing algorithm was a compressed version of the input audio signal.

A preferred audio encoding method for generating an encoded audio signal includes an encoding step 400, an intermediate audio signal 195 in accordance with a first encoding algorithm, a first encoding algorithm based on an information model of hearing and generation, spectral information representing the first output signal is encoded audio signal; step 500 is the encoding of the audio of the intermediate signal 195 in accordance with the second coding algorithm, which is based on the information model of the source and generation, in the second output signal, the parameters of the information model of the source, representing the intermediate signal 195, and the step of the general preliminary processing 100 of the input audio signal 99 are encoded in order to obtain the audio intermediate signal 195, where in the general pre-processing step, the input audio signal 99 is processed so that the audio intermediate signal 195 is a compressed version of the input audio signal 99, wherein the encoded audio signal includes, for a specific part of the audio signal, either a first output signal or a second output signal. The method preferably includes a further step encoding a certain part of the intermediate audio signal using the first encoding algorithm, or using the second encoding algorithm, or the encoding signal using both algorithms, generating an output signal as the result of both algorithms, or the result of the first encoding algorithm, or the result of the second coding algorithm.

In general, the audio encoding algorithm used in the first encoding channel 400 reflects and models the audio perception situation. The hearing organ for audio information is usually the human ear. The human ear can be modeled as a frequency analyzer. Therefore, the output of the first coding channel encodes spectral information. Preferably, the first coding channel further includes a physical-acoustic model in order to further use a masking physical-acoustic threshold. This masking physico-acoustic threshold is used when quantizing audio spectral components, where preferably the quantization is performed such that quantization noise is introduced when quantizing spectral audio components that are hidden below the masking physico-acoustic threshold.

The second coding channel is an information source model that reflects the formation of audio sound. Therefore, source information models can include a speech model that is reflected by the LPC analytic step, that is, the step of converting the signal over a time interval to the LPC region and subsequently processing the residual LPC signal, i.e., the drive signal. Alternative sound source models are sound models for representing a specific instrument or any other sound generators, such as a specific sound source existing in the real world. The choice between different sound source models can be made when several sound source models are available, for example, the choice can be based on SNR calculation, that is, on the calculation which of the source models selects the best one suitable for encoding a certain time part / or frequency part of the audio signal. Preferably, however, the switching between coding channels is performed in a time interval, that is, that a certain time interval is encoded using one model, and a certain other time interval of the intermediate signal is encoded using another encoding channel.

Information source models are represented by certain parameters. With respect to the speech model, the parameters are the LPC parameters and the encoded excitation parameters when considering a modern speech encoder such as AMR-WB +. AMR-WB + includes an ACELP encoder and a TLC encoder. In this case, the encoded drive parameters may be global transmission, noise level, and variable length codes.

Fig. 3b illustrates a decoder corresponding to the encoding device illustrated in Fig. 3a. In general, FIG. 3b illustrates an audio decoder for decoding an encoded audio signal to obtain a decoded audio signal 799. The decoder includes a first decoding channel 450 to decode a signal encoded in accordance with a first encoding algorithm based on a hearing information model. The audio decoder further includes a second decoding channel 550 in order to decrypt the encoded information message in accordance with a second encoding algorithm based on the model of the information source. The audio decoder further includes a combiner for combining the output signals of the first decoding channel 450 and the second decoding channel 550 to obtain a combined signal. The combined signal illustrated in FIG. 3b is an audio decoded intermediate signal 699 that is supplied to a common post-processing unit processing the decoded audio intermediate signal 699, which is a combined signal generated by combiner 600 so that the output of the common preprocessing unit is an extended version combined signal. Thus, the decoded audio signal 799 has an expanded information content compared to the decoded audio intermediate signal 699. This information extension is provided by a common post-processing unit using pre / post processing parameters that can be transmitted from the encoder to the decoder or which can be received from the decrypted audio intermediate signal directly. Preferably, however, the pre / post processing parameters are transmitted from the encoder to the decoder, as this procedure allows improved quality of the decoded audio signal.

FIG. 3c illustrates an audio encoder for encoding an input audio signal 195, which may be equal to the intermediate audio signal 195 in FIG. 3a in accordance with a preferred solution of the present invention. The audio input signal 195 is present in the first region, which may, for example, be a time interval, but which may also be any other region, such as a frequency region, an LPC region, an LPC spectral region, or any other region. In general, conversion from one region to another region is performed by a conversion algorithm, such as any of the known conversion time / frequency algorithms or conversion frequency / time algorithms.

An alternative conversion from a time slot, for example, to an LPC region, is the result of LPC filtering of the time slot signal, which results in a difference LPC signal or an excitation signal. Any other filtering operations that generate a filtered signal that affects a significant number of signal samples before conversion can be used as a conversion algorithm, depending on the circumstances. Therefore, weighting an audio signal based on an LPC weighting filter is a further transform that generates a signal in the LPC domain. In a time / spectral transformation, a change in a single spectral component affects all components of the time interval before the conversion. Similarly, the modification of any sample of the time interval will affect each component of the frequency domain. Similarly, modifying the excitation signal sample in a situation with the LPC region will, due to the length of the LPC filter, have an effect on a significant number of components before filtering the LPC. Similarly, modifying a component before LPC conversion will affect many of the components obtained by this LPC conversion due to the internal LPC filter memory effect.

The audio encoder of FIG. 3c includes a first encoding channel 400 that generates a first encoded signal. This first encoded signal may be in the fourth region, which is, in a preferred solution, the time-spectral region, that is, the region that is obtained when the time-slot signal is processed through a time / frequency conversion.

Therefore, in the first encoding channel 400 for encoding an audio signal, a first encoding algorithm is used to obtain a first encoded signal, where this first encoding algorithm may or may not include a time / frequency conversion algorithm.

The audio encoder further includes a second encoding channel 500 in order to encode the audio signal. In the second coding channel 500, a second coding algorithm is used that is different from the first coding algorithm to obtain a second encoded signal.

The audio encoder further includes a first switch 200 so as to switch between the first encoding channel 400 and the second encoding channel 500 so that for a portion of the input audio signal, either the first encoded signal at the output of block 400 or the second encoded signal at the output a second encoding channel has been included in the output of the encoder.

Thus, when for a certain part of the input audio signal 195, the first encoded signal in the fourth region is included in the output of the encoder, the second encoded signal, which is either the first processed signal in the second region or the second processed signal in the third region, is not included in the output encoder signal. This ensures that this encoder has an effective bit rate. In solutions, any time slots of an audio signal that are included in two different encoded signals are small compared to the frame length, as will be discussed in connection with FIG. 3. These small parts are useful for cross-fading one coded message signal with another coded signal in the case of a switch switch to reduce distortion that might occur without a cross-fade. Therefore, in addition to the interval of smooth overlapping signals, each block of the time interval is represented by the encoded signal of only one area.

As illustrated in FIG. 3c, the second coding channel 500 includes a converter 510 in order to convert the audio signal in the first region, i.e., signal 195, into the second region. In addition, the second coding channel 500 includes a first processing channel 522 in order to process the audio signal in the second region to obtain a first processed signal, which is preferably also in the second region so that the first processing channel 522 does not change the region .

The second coding channel 500 further includes a second processing channel 523, 524, which converts the audio signal in the second region to a third region, which is different from the first region and which also differs from the second region, and which processes the audio signal in the third region to get the second processed signal at the output of the second processing channel 523, 524.

In addition, the second coding channel includes a second switch 521 in order to switch between the first processing channel 522 and the second processing channel 523, 524 so that for a portion of the input audio signal into the second coding channel or the first processed signal in the second region or the second processed signal in the third area were in the second encoded signal.

Fig. 3d illustrates a corresponding decoder in order to decode the encoded audio signal generated by the encoder in Fig. 3c. In general, each block of the audio signal of the first region is represented by a signal of the second region or a signal of the third region or an encoded signal of the fourth region, except, possibly, a smooth overlap interval, which is preferably small compared to the length of one frame, to obtain a system that is how much this is possible at the critical limit of the sampling rate. The encoded audio signal includes a first encoded signal, a second encoded signal in a second region and a third encoded signal in a third region, wherein the first encoded signal, the second encoded signal and the third encoded signal all relate to different time portions of the decoded audio signal, and wherein the second region, the third region and the first region for the decoded audio signal are different from each other.

The decoder includes a first decoding channel in order to decrypt a signal based on the first coding algorithm. The first decoding channel is illustrated by blocks 431, 440 in FIG. 3d and preferably includes a frequency / time converter. The first encoded signal is preferably in the fourth region and is converted to the first region, which is the region for the decoded output signal.

The decoder in FIG. 3d also includes a second decoding channel, which includes several elements. These elements are the first reverse processing channel 531 to reverse transform the second encoded signal and obtain the reverse processed signal in the second region at the output of block 531. The second decoding channel further includes a second reverse processing channel 533, 534 for reverse processing the third encoded signal so that receive the second back-processed signal in the second region, where the second channel of the reverse processing includes a converter in order to convert the signal from the third region to the second area.

The second decoding channel also includes a first combiner 532 for combining the first back-processed signal and the second back-processed signal to obtain a signal in the second region where this combined signal, at the first time, is only influenced by the first back-processed signal and, at a later point in time, only under the influence of a second back-processed signal.

The second decoding channel also includes a converter 540 in order to convert the combined signal into a first region.

Finally, the decoder illustrated in FIG. 3d includes a second combiner 600 in order to combine the first decoded signal from blocks 431, 440 and the output of converter 540 to obtain a decoded output in the first region. Further, the decoded output signal in the first region at the first time is only influenced by the signal generated by the converter 540, and at a later time, it is only influenced by the first decoded signal generated by blocks 431,440.

This situation is illustrated from the point of view of the encoder shown in FIG. The upper part of FIG. 3e illustrates in a schematic representation the audio signal of a first region, such as an audio signal of a time interval, where the time index increases from left to right, and diagram 3 could be considered as a stream of audio samples representing the signal 195 in FIG. 3c. Fig. 3e illustrates frames 3a, 3b, 3c, 3d, which can be generated by switching between the first encoded signal and the first processed signal and the second processed signal, as illustrated in diagram 4 in Fig. 3e. The first encoded signal, the first processed signal, and the second processed signal are in different areas, and to make sure that the switch between different areas does not cause distortion on the side of the decoder, the frames of the signal of the time interval have a smoothly overlapping range, which is indicated by the area crossfade, and such a crossfade area is shown in frames 3b and 3c. However, there is no crossfade area between the 3d frames and 3c, which means that the 3d frame is also represented by the second processed signal, that is, the signal in the third area, and there is no change in the area between the 3c and 3d frames. Therefore, in general, it is preferable to set the crossfade area when there is no change in the area, and set the crossfade area, i.e. the time interval of the audio signal, which is encoded using two encoded / processed signals, when there is a change in the area, i.e. there is a switch of either of the two switches. Preferably, the crossfade is implemented for other area changes.

In a solution in which a first encoded signal or a second processed signal was generated by an MDCT procedure, having, for example, 50 percent overlap, of each sample of a time interval included in two subsequent frames. Due to the nature of MDCT, however, this does not lead to overflow, since MDCT is a critically sampled system. In this context, a critically sampling system means that the number of spectral components is the same as the number of samples in a time interval. MDCT is advantageous in that the intersection effect is provided without a defined intersection area, so that the intersection of the MDCT block and the next MDCT block is provided without overflow, which would violate the critical requirement for sampling.

Preferably, the first coding algorithm in the first coding channel of the department is based on the hearing information model, and the second coding algorithm in the second coding channel is based on the information source model or SNR model. The SNR model is definitely not associated with a specific sound generation mechanism, but it is one coding method that can be selected from a variety of coding methods based, for example, on solving a closed loop. Thus, the SNR model is any available coding model, but which does not have to be related to the physical nature of the sound generator, which is any parameterized coding model that differs from the hearing information model, which can be chosen by solving a closed loop and especially by comparing different SNR results of various models.

FIG. 3c shows the controller 300, 525. This controller may include the functionality of the selection unit 300 shown in FIG. 1a, and, optionally, may include control device functionality with a switch 525 shown in FIG. 1a. In general, the controller is designed to control the first switch and the second switch on an adapted signal path. The controller serves to analyze the input signal of the first switch, or the output signal of the first or second encoding channel, or the output signals obtained by encoding and decoding in the first and second encoding channels using the target function. Alternatively, or additionally, the controller serves to analyze the input signal to the second switch or the output signal of the first processing channel or the second processing channel or a signal obtained by processing and reverse processing from the first processing channel and the second processing channel using the target function.

In one solution, in the first coding channel or in the second coding channel, the results of a time / frequency conversion algorithm such as MDCT or the MDST algorithm are combined, which differs from the direct FFT transformation, in which the matching effect is not implemented. In addition, one or both channels include a quantizer / entropy encoding unit. Specifically, only the second processing channel of the second encoding channel includes a time / frequency converter introducing a combining operation, and the first processing channel of the second encoding channel includes a quantizer and / or entropy encoder and does not perform a combining operation. The combining operation is carried out by a time / frequency converter, preferably using processing by an analytical window, and an MDCT transformation algorithm. Specifically, processing by the analytic window serves to apply the window function to successive overlapping frames so that the sample signal processed by the window function is in at least two subsequent frames processed by the window function.

In one solution, the first processing channel includes an ACELP encoder, and the second processing channel includes an MDCT spectral converter and a quantizer in order to quantize the spectral components and obtain quantized spectral components, where each quantized spectral component is zero or is determined by one quantization index of the set of different possible quantization indices .

In addition, it is preferred that the first switch 200 operates in open loop mode and the second switch operates in closed loop mode.

As stated previously, both coding channels serve to encode the audio signal in the block in an intelligent way in which the first switch or the second switch is switched so that the switch takes place, at a minimum, after the block of a predetermined number of samples of the signal, a predetermined number forming the frame length for corresponding switch. Thus, the interval in order to switch the first switch may be, for example, a block of 2048 or 1028 samples, and the frame length based on the switching of the first switch 200, and may be variable, but preferably fixed for such a rather long period.

And vice versa, the block size for the second switch 521, that is, when the second switch 521 switches from one method to another, is substantially smaller than the block size for the first switch. Preferably, both block sizes for the switches are selected such that the dyne of the longer block is an integer times the size of the shorter block. In a preferred solution, the block size of the first switch is 2048 or 1024, and the block size of the second switch is 1024 or more preferably 512 and even more preferably 256 and even more preferably 128 samples so that the second switch can switch up to 16 times when the first switch switches only once. The preferred maximum block size ratio is 4: 1.

In yet another embodiment, the controller 300, 525 serves to perform the separation of speech and music for the first switch in such a way that the choice of speech prevails relative to the choice of music. This decision made the choice of speech, even when part of less than 50% of the frame for the first switch is speech and part of more than 50% of the frame is music.

In addition, the controller serves to already switch to speech mode when a rather small part of the first frame is speech and, specifically, when a part of the first frame is speech, which is 50% of the length of the smaller second frame. Thus, preferably, the speech / approval switching solution is already switched to speech, even when, for example, only 6% or 12% of the block corresponding to the frame length of the first switch is speech.

This procedure is preferable in order to fully utilize in one solution the ability to save the bit rate of the first processing channel, which has a coding core with a voice of speech, and not lose quality for the rest of the large first frame, which is not speech due to the fact that the second processing channel It includes a converter and, therefore, is useful for audio signals that also have non-speech signals. Preferably, this second processing channel includes an overlapped MDCT transform that is critically selected and which, even with small window sizes, provides high efficiency and is free from alignment by canceling overlapping alignment processing, such as overlapping and adding on the side of the decoder. In addition, the large block size for the first coding channel, which is preferably an AAC-like MDCT coding channel, is useful since non-speech signals are usually fairly constant, and the long conversion window provides high-frequency resolution and, therefore, high quality and, in addition, low bit rate thanks to the psychoacoustically controlled quantization module, which can also be applied to the transformation based on the conversion mode in the second processing channel of the second coding channel Ania.

Regarding the decoder illustrated in FIG. 3d, it is preferred that the transmitted signal includes an explicit indicator, such as third-party information 4a, as illustrated in FIG. 3e. This third-party information 4a is extracted by a bitstream recognizer not illustrated in FIG. 3d to direct the corresponding first encoded signal, the first processed signal, or the second processed signal to the correct processor, such as a first decoding channel, a first reverse processing channel, or a second reverse processing channel, depicted in fig.3d. Therefore, the encoded signal has not only an encoded / processed signal, but also includes third-party information regarding these signals. In other solutions, however, there may be implicit signaling that allows the bitstream analyzer of the decoder side to distinguish certain signals. This is described generally in FIG. 3e, where the first processed signal or the second processed signal is the output of the second coding channel and, therefore, the second encoded signal.

Preferably, the first decoding channel and / or the second reverse processing channel includes an MDCT procedure to transform the spectral region into a time interval. For this purpose, an overlapping adder is installed that performs the function of canceling time intervals, which, at the same time, provides a crossfade to avoid blocking distortions. In general, the first decoding channel converts the signal encoded in the fourth region to the first region, while the second reverse processing channel converts the third region to the second region, and the converter subsequently coupled to the first combiner converts the second region to the first region so that at the input of combiner 600, there were only signals of the first region, which represent the decrypted output signal of the solution shown in fig.3d.

Figs. 4a and 4b illustrate two different solutions that differ in the location of the switch 200. In Fig. 4a, a switch 200 is placed between the output of the common preprocessing unit 100 and the input of two coding channels 400, 500. The solution in Fig. 4a provides an audio signal only in the only coding channel, while the other coding channel, which is not connected to the output signal of the common pre-processing unit, does not work and, therefore, is turned off or is in a state of sleep. This solution is preferable in that the inactive coding channel does not consume energy and computational resources, which is useful for mobile applications, in particular for applications that are battery powered and therefore have a general limitation of energy consumption.

On the other hand, however, the solution in FIG. 4b may be preferred when power consumption is not a problem. In this solution, the coding channels 400, 500 are active all the time, and only the output signal of the selected coding channel for a certain part of the time and / or a certain part of the frequencies is sent to the bitstream by a shaper, which can be implemented as a multiplexer of the bitstream 800. Therefore, in the solution in FIG. 4b, the coding channels are active all the time, and the output of the coding channel that is selected by the decision unit 300 is input to the bit output stream, while the output, that is, the output signal of another non-selected coding channel 400, is rejected, that is, this data does not enter the output bitstream and the encoded audio signal.

Preferably, the second encoding rule / decoding rule is based on the LPC coding algorithm. In LPC-based speech coding, a separation is made between quasiperiodic pulse-like excitation signal segments or signal parts and noise-like excitation signal segments or signal parts. This is done to achieve a very low LPC bitrate by vocoders (2.4 kbps) as shown in FIG. 7b. However, with an average bit rate in CELP encoders, excitation is generated to add scaled vectors from the adaptive code table and fixed code table.

Quasiperiodic, pulse-like excitation signal segments, that is, signal segments having a specific pitch, are encoded using other algorithms than noise-like excitation signals. While quasiperiodic impulse-like excitation signals are associated with voice-enabled speech, signals similar to noise are associated with non-voice-based speech.

Figures 5a and 5d show an example. Here, quasi-periodic, pulse-like signal segments or signal portions and noise-like signal segments or signal portions discussed as an example are presented. Specifically, the voice-enabled speech presented in FIGS. 5c and 5d, as illustrated in FIG. 5a in the time domain and in FIG. 5b in the frequency domain, is discussed as an example of a quasiperiodic, pulse-like part of the signal, and the voice-free speech segment is discussed as an example of a part of a signal similar to noise. In general, speech can be classified as having a voice or voice, not having a voice or not voice and mixed. Charts in the areas of time and frequency for the selected voice and non-voice segments are shown in figa 5d. Voice speech is quasiperiodic in the time interval and harmonically structured in the frequency domain, while the speed of non-voice speech is similar to a random broadband signal. The spectrum of voice speech over a short time interval is characterized by a good formant structure. A good harmonic structure is a consequence of the quasiperiodicity of speech and can be attributed to vibrating vocal cords. The formant structure (spectral envelope) arises from the interaction of the source and vocal tract. The vocal tract consists of the pharynx and the cavity of the mouth. The shape of the spectrum envelope, which "corresponds" to a spectrum with a speech voice for a short period of time, is associated with the characteristics of the transmission of the vocal tract and the spectral tilt (6 decibels / Octave) due to the glottal pulse. The envelope of the spectrum is characterized by a number of peaks, which are called formants. The formants are the resonant modes of the vocal tract. For the middle vocal tract within 5 kHz, there are three to five formants. The amplitudes and positions of the first three formants, usually found below 3 kHz, are quite important both in speech synthesis and perception. Higher-frequency formants are also important for a wide group and non-voice representations of speech. The properties of speech are associated with the physical speech forming system as follows. Voice speech is generated by excitation of the vocal tract with a quasiperiodic glottal air pulse created by vibrating vocal cords. The heart rate is referred to as the fundamental frequency or pitch. Non-voice speech is formed by compressing air as it passes through a vocal tract. Nasal sounds are due to the acoustic connection of the nasal tract with the vocal tract, and consonant sounds are formed by the release of air under pressure, which was created behind the obstruction in the tract.

Thus, the noise-like part of the audio signal shows neither a pulse-like structure in the time interval, nor a harmonic structure in the frequency domain, as illustrated in FIG. 5c and FIG. 5d, which differs from the quasiperiodic pulse-like part, as illustrated, for example, in figa and fig.5b. As will be described in general terms later, separation between the noise-like parts and the quasiperiodic pulse-like parts can also occur after the LPC for the excitation signal. In the LPC method, the vocal tract is modeled and the excitation of the vocal tracts is extracted from the signal.

In addition, quasiperiodic pulse-like parts and noise-like parts can occur in a timely manner, that is, which means that part of the audio signal is noisy at the same time, and the other part of the audio signal is quasiperiodic, i.e. tonal. Alternatively, or additionally, the signal feature may differ in different frequency ranges. Thus, determining whether an audio signal is noise or tonal can also be performed with frequency selection so that a certain frequency range or several specific frequency ranges are considered noise and other frequency ranges are considered tonal. In this case, a specific time portion of the audio signal could include tonal components and noise components.

Fig. 7a illustrates a linear model of a speech forming system. This system assumes a two-stage excitation, that is, a driving pulse of voice speech, as shown in FIG. 7c, and random noise for non-voice speech, as shown in FIG. 7d. The vocal tract is modeled as an ideal filter 70, which processes the pulses shown in FIG. 7c or FIG. 7d generated by the glottal model 72. Therefore, the system in FIG. 7a can be reduced to the ideal filter model in FIG. 7b having an amplification unit 77, the direct path 78 and the feedback path 79 and the addition unit 80. On the feedback path 79, there is a predictive filter 81, and the whole source model synthesis system illustrated in FIG. 7b can be represented using the z-region function as follows way:

S (z) = g / (1-A (z)) X (z),

where g represents the gain, A (z) is the predictive filter, as determined by LP analysis, X (z) is the excitation signal, and S (z) is the output synthesized speech.

Figs and 7d give a graphical description of the time interval with synthesized voice and non-voice speech using a linear source system model. This system and excitation parameters in the above equation are unknown and must be determined from a finite set of speech samples. The coefficients A (z) are obtained using linear prediction of the input signal and quantization of the filter coefficients. In a p-th order linear predictor, the current speech sequence pattern is predicted as a linear combination of p transmitted samples. The predictor coefficients can be determined by known algorithms, such as the Levinson-Durbin algorithm, or in general by the method of autocorrelation or reflection method.

Fig. 7e illustrates a more detailed description of the analytical LPC unit 510. The audio signal is included in the filter parameter determination unit, which determines the filter information A (z). This information is created as short-term prediction information required for a decoder. Short-term prediction information is requested by the actual prediction filter 85. The current sample of the audio signal and the expected value for the current sample are supplied to the subtractor 86 and subtracted so that for the current sample the prediction error signal is generated on line 84. The sequence of such errors of the prediction signal samples is very schematically illustrated in figs or 7d. Therefore, the diagrams in figa, 7b can be considered as a kind of corrected similar to a pulse signal.

While FIG. 7e illustrates a preferred method for calculating a drive signal, FIG. 7f illustrates a preferred method for calculating a weighted signal. Unlike FIG. 7e, the filter 85 differs when γ differs from 1. A value less than 1 is preferred for γ. In addition, in the present block, 87 µ is a number, preferably less than 1. In general, the elements in FIGS. 7e and 7f can be implemented as described in 3GPP TS 26.190 or 3GPP TS 26.290.

FIG. 7G illustrates reverse processing, such as in element 537 in FIG. 2b, which can be applied on the decoder side. In particular, block 88 generates an unweighted signal from the weighted signal, and block 89 calculates the excitation from the unweighted signal. In general, all signals except the non-weighted signal in FIG. 7G are in the LPC region, but the drive signal and the weighted signal are different signals in the same region. Block 89 generates an excitation signal that can be used in conjunction with the output of block 536. Then, a common inverse LPC transform can be performed in block 540, shown in FIG. 2b.

Subsequently, the synthesis synthesis coding apparatus CELP shown in FIG. 6 to illustrate modifications related to this algorithm. The CELP encoder is discussed in detail in "Speech Coding: A Tutorial Review", Andreas Spanias, Proceedings of the IEEE, Vol. 82, No.10, October 1994, pages 1541-1582. The CELP encoder, as illustrated in FIG. 6, includes a long-term predictor 60 and a short-term predictor 62. In addition, a code table that is designated 64 is used. The perceptual weighting filter W (z) is represented by block 66, and the error minimization controller is indicated by block 68 . The signal s (n) is an input signal on a time interval. Being perceptually weighted, the weighted signal is supplied to a subtractor 69, which calculates the error between the weighted synthesized signal at the output of block 66 and the original weighted signal Sw (n). In general, the short-term prediction filter coefficients A (z) are calculated by the LP analysis unit and quantized into A (z), as indicated in FIG. 7e. The long-term prediction information Af (z) includes the long-term gain prediction (transmission) g and the vector quantized index, i.e., the code table references to the prediction error signal in the output of the analysis LPC block marked as block 10a in FIG. 7e are calculated. The LTP parameters are pitch attenuation and gain (transmission). In CELP, this is usually implemented as an adaptive code table containing the past excitation signal (non-differential). CB adaptive attenuation and gain are found by minimizing the mean-square weighted error.

The CELP algorithm then encodes the difference signal obtained after short-term and long-term predictions using a code table, for example, Gaussian sequences. The ACELP algorithm, where "A" stands for "Algebraic", has a specific algebraically designed code table.

The code table may contain more or less long vectors, where each vector is several samples long. The gain g scales the code vector, and the resulting code is filtered by a long-term prediction synthesis filter and a short-term prediction synthesis filter. The “optimal” code vector is chosen so that the perceptually weighted mean square error at the output of the subtractor 69 is minimized. The search process in CELP is done by optimizing the analysis through synthesis as illustrated in FIG. 6.

For specific cases, when the frame is a mixture of non-voice and voice speech or when speech is formed from music, TLC coding designed to encode excitation in the LPC region may be more appropriate. The TLC coding procedure uses a weighted signal in the frequency domain without assuming excitation to form. TLC is then more natural than CELP coding and is not limited to voice or non-voice source excitation models. TLC is also source model-centric coding using a linear predictive filter in order to simulate formants of signals like speech.

AMR-WB + -like encoders have a choice between different TLC and ACELP modes, as is known from the AMR-WB description. TLC modes differ in the length of the discrete Fourier transform block (DFT = ORT) for different modes, and the best mode can be selected using the analysis through synthesis approach or through the direct “lead” mode.

As discussed in connection with FIGS. 2a and 2b, the common preprocessing unit 100 preferably includes an integrated multi-channel unit (surround / integrated stereo device) 101 and an additional extension unit of the frequency band 102. Accordingly, the decoder includes the extension unit of the frequency band 701 and a series-connected multi-channel block 702. Preferably, in the encoder, the combined multi-channel block 101 is connected before the band extension block 102, and, on the decoder side, the band expansion block 701 Frequency s must be connected in front of the combined multi-channel block 702 with respect to the direction of signal processing. Alternatively, however, the common preprocessing unit may include an integrated multi-channel unit without a series-connected frequency band extension unit, or a frequency extension unit without a connected integrated multi-channel unit.

A preferred example for a combined multi-channel block on the side of the encoder 101a, 101b and on the side of the decoder 702a, 702b is illustrated in FIG. Many of the original input channels E enter the mixer with the abbreviation of channels 101a so that the mixer with the abbreviation forms k transmission channels, where the number k is greater than or equal to one and less than or equal to E.

Preferably, the input channels E enter a combined multi-channel parameter analyzer 101b, which generates parameter information. This parameter information is preferably encoded by a method without loss of information (entropy coding), such as difference coding and subsequent coding using the Huffman algorithm or, alternatively, subsequent arithmetic coding. The encoded parameter information generated by block 101b is transmitted to the parameter decoder 702b, which may be part of block 702 in FIG. 2b. The parameter decoder 702b decodes the transmitted parameter information and transmits the decoded parameter information to a mixer with extension 702a. A channel expansion mixer 702a receives k transmission channels and generates multiple L channels at the output, where the number L is greater than or equal to k and less than or equal to E.

Information about the parameters may include inter-channel level differences, inter-channel temporal differences, inter-channel phase differences and / or inter-channel differences of coherence measures, as is known in the BCC technique, or as is known and described in detail in the MPEG standard environment. The number of transmitted channels may be the only mono channel for ultra-low bitrate applications or may include a compatible stereo application or may include a compatible stereo signal, that is, two channels. Typically, the number of input channels E may be five or possibly even more. Alternatively, the plurality of input channels E may also be a plurality of audio objects E, as is known in the context of encoding spatial audio objects (SAOC).

In one solution, the mixer abbreviates the weighted or non-weighted addition of the original input channels E, or addition E of the input audio objects. In the case of audio objects, such as input channels, the combined multi-channel parameter analyzer 101b calculates the object's audio parameters, such as a correlation matrix between audio objects, preferably for each time interval and even more preferably for each frequency range. To this end, the entire frequency range can be divided into at least 10 and, preferably, 32 or 64 frequency ranges.

FIG. 9 illustrates a preferred solution for executing the bandwidth extension unit 102 in FIG. 2a and the corresponding bandwidth extension unit 701 in FIG. 2b. On the encoder side, the bandwidth expansion unit 102 preferably includes a low-pass filter 102b, a lower-sampler sampler unit that follows the low-pass filter, or which is part of the inverse QMF that acts on only half of the QMF bands, and a high-frequency analyzer 102a. The original input audio signal to the bandwidth extension unit 102 is filtered by a low-pass filter to form a signal in the low frequency band, which is supplied to the coding channel and / or to the switch. The low-pass filter has a cutoff frequency that can range from 3 kHz to 10 kHz. In addition, the bandwidth extension unit 102 includes a high-frequency analyzer in order to calculate the bandwidth extension parameters, such as spectral envelope parameter information, noise level parameter information, reverse filtering parameter information, further parameter information regarding certain harmonic lines in the high-frequency band and additional parameters discussed in detail in the MPEG-4 standard in the chapter related to the repetition of the frequency range.

On the decoder side, the band extension unit 701 includes a reducing agent 701 a, a regulator 701 b, and a combiner 701 c. Combiner 701c combines the decoded low-frequency signal and the reconstructed and adapted high-frequency signal generated by the adjuster 701b. The input to the adjuster 701b is provided with a controlled reducer to obtain a high frequency signal from a low frequency signal, by repeating the range or, in general, by expanding the frequency band. The corrections made by the reducing agent 701a may be corrections made in a harmonic or non-harmonic way. The signal generated by the reducing agent 701a is subsequently adapted by the controller 701b using the transmitted parametric information about the extension of the frequency band.

As indicated in FIG. 8 and FIG. 9, in a preferred solution, the described units may have a mode control input. This mode control input signal is obtained from the output of the decision unit 300. In such a preferred solution, the parameter of the corresponding unit can be adapted to the output of the selection unit, that is, whether the choice of speech or the choice of music for a specific time portion of the audio signal is made in the preferred solution. Preferably, the mode control refers only to one or more of the functionalities of these blocks, but not to all the functionalities of these blocks. For example, the selection may affect only the reducing agent 701a, but may not affect the other blocks in FIG. 9, or may, for example, affect only the combined multi-channel parameter analyzer 101b in FIG. 8, but not the other blocks in FIG. . This embodiment is preferable, as this results in higher flexibility, higher quality and lower bitrate of the output signal by providing flexibility to the common preprocessing unit. On the other hand, however, the use of algorithms for both types of signals in a common pre-processing unit allows for an efficient encoding / decoding scheme.

10a and 10b illustrate two different embodiments of a selection unit 300. FIG. 10a shows an open loop solution. Here, the signal of the analyzer 300a of the decision block obeys certain rules to decide whether a certain time part or a certain frequency region of the input signal has a feature that requires that this part of the signal be encoded in the first encoding channel 400 or in the second encoding channel 500. C for this purpose, the signal analyzer 300a may analyze the input audio signal to the common pre-processing unit or may analyze the audio signal generated by the general pre-processing stage and, that is, the audio intermediate signal or may analyze an intermediate signal in the common pre-treatment unit, such as a mixer output signal with a reduction which may be mono signal or which may be a signal having k channels 8. On the output side, the signal analyzer 300a generates a switching decision in order to control the switch 200 on the encoder side and the corresponding switch 600 or combiner 600 on the decoder side.

Although the second switch 521 is not discussed in detail, it must be emphasized that the second switch 521 can be positioned in a manner similar to the positioning of the first switch 200, as discussed in connection with FIGS. 4a and 4b. Thus, an alternative position of the switch 521 in FIG. 3c is the output of both processing channels 522, 523, 524 so that the processing channels work in parallel, and only the output signal of one processing channel is recorded in the bit stream through a bitstream generator, which is not illustrated in figs.

In addition, the second combiner 600 may have certain crossfade functionality, as discussed in FIG. 4c. Alternatively or additionally, the first combiner 532 might have the same crossfade functionality. In addition, both combiners may have the same crossfade functionality, or may have different crossfade functionality, or may not have any crossfade functionality at all, so both combinators will be switches without any additional crossfade functionality.

As discussed previously, both switches can be controlled by solving an open loop or solving a closed loop, as discussed in connection with FIGS. 10a and 10b, where the controllers 300, 525 in FIG. 3c may have different or the same functionality for both switches .

In addition, time distortion functionality that is adaptive to the signal may exist not only in the first encoding channel or in the first decoding channel, but may also exist in the second processing channel of the second encoding channel on the encoder side as well as on the decoder side. Depending on the processed signal, both time distortion functionalities may have the same time distortion information so that the same time distortion is applied to the signals in the first region and in the second region. This reduces the processing load and can be useful in some cases, in cases where consecutive blocks have the same time distortion characteristics. In alternative solutions, however, it is preferable to have independent time distortion evaluators for the first coding channel and the second processing channel in the second coding channel.

The audio signal encoded according to the invention may be stored on a digital storage medium or may be transmitted in a transmission medium, such as a wireless transmission medium or a transmission medium by wire, such as the Internet.

In various solutions, the switch 200 shown in FIGS. 1a or 2a switches between two encoding channels 400, 500. In a further solution, there may be additional encoding channels, such as a third encoding channel, or even a fourth encoding channel, or even more encoding channels. On the decoder side, a switch 600 shown in FIG. 1b or 2b switches between two decoding channels 431, 440 and 531, 532, 533, 534, 540. In a further solution, there may be additional decoding channels, such as a third decoding channel, or even fourth decoding channel, or even more decoding channels. Similarly, other switches 521 or 532 may switch more than between two different coding algorithms when such additional encoding / decoding channels are present.

12A illustrates a preferred embodiment of an encoder, and FIG. 12B illustrates a preferred embodiment of a corresponding decoder. In addition to the elements previously discussed with respect to the respective reference numbers, the solution in FIG. 12A illustrates a separate physico-acoustic module 1200, and further illustrates a preferred embodiment of additional encoder tools illustrated in block 421 in FIG. 11A. These additional tools are Temporary Noise Shaper (TNS) 1201 and Mid / Side Encoding Tool (MYS) 1202. In addition, the additional functionality of elements 421 and 524 is illustrated in block 421/542 as a combined execution of scaling, noise filling analysis, quantization, and arithmetic coding. spectral components.

In a corresponding embodiment of the decoder of FIG. 12B, additional elements are illustrated, which are the decryption tool M \ S 1203 and the tool of the TNS decoder 1204. In addition, a bass postfilter not illustrated earlier is indicated as 1205. The processing unit by the transition window function 532 corresponds to element 532 on figv, which is illustrated as a switch, but which performs a kind of crossfade, which may be a crossfade with an increased sampling frequency or a crossfade with a critically selected sampling frequency. The latter is carried out as an MDCT operation, where the signals at two overlapping time intervals overlap and add up. Where possible, processing with a critically selected sampling rate is preferably used, since the overall bitrate can be reduced without loss of quality. An additional processing unit with a transition window function 600 corresponds to a combiner 600 in FIG. 2B, which is again illustrated as a switch, but it is clear that this element performs a kind of crossfade with a critically selected sampling rate, or with an uncritically selected sampling frequency, to avoid blocking distortions, and specific distortions that occur when switching, when one block was processed in the first channel, and another block was processed in the second channel. When, however, the processing in both channels is in good agreement with each other, then the crossfade operation may be “worse” than hard switching, where the crossfade, as is understood, is a “soft” switching between both channels.

The concept illustrated in FIGS. 12A and 12B allows encoding signals having an arbitrary combination of speech and audio, and this concept demonstrates comparable or better than the best encoding technology that could be created for encoding or speech or arbitrary audio content. The general structure of the encoder and decoder can be described as general pre-post processing, consisting of a functional unit of MPEG surround (MPEGS), for control with stereo or multi-channel processing, and an extended unit of SBR (eSBR), which controls the parametric representation of higher audio frequencies in the input signal. Then, there are two channels: one consisting of a modified advanced audio coding instrument (AAC), and the other consisting of linear prediction coding (LP or LPC region), which, in turn, is either a representation of a frequency domain or a representation of a temporal interval LPC residual (differential) signal. All transmitted spectra for both AAC and LPC are presented in the MDCT domain after quantization and arithmetic coding. The time slot representation uses the ACELP coding excitation scheme. The basic structure is shown in FIG. 12A for the encoder and FIG. 12B for the decoder. The data flow in this diagram is directed from left to right, from top to bottom. The function of the decoder is to search for a description of a quantized audio spectrum or a temporal representation of a signal in a bit stream, and to decode the quantized quantities and other reconstruction information.

In the case of transmission of spectral information, the decoder must restore the quantized spectra, and perform the process of spectrum reconstruction using any tools in the bitstream in order to obtain the actual spectrum of the signal, as described in the input bitstream, and finally convert the spectrum from the frequency domain to time interval. After the initial reconstruction and scaling of the reconstructed spectrum, there are additional tools that modify one or more spectra to provide more efficient coding.

In the case of transmitting the time representation of the interval signal, the decoder must reconstruct the quantized time signal and carry out the process of restoring the time signal using any tools in the bitstream to obtain the actual signal in the time interval, as described in the input bitstream.

For each of the additional instruments that affect the signal data, it is possible to “pass through”, and in all cases when processing is omitted, the spectra or time samples at the input are transmitted directly through the instrument without modification.

In places where the bitstream changes its representation of the signal from the time domain to the spectral region or from the LP region to a non-LP region or vice versa, the decoder should facilitate the transition from one region to another by appropriately processing the overlay-add transition window function.

The processing of eSBR and MPEGS is applied similarly to both coding paths after transition processing.

The input signal to the bit demultiplexing tool A stream is a bitstream. The demultiplexer divides the bitstream into parts for each tool, and provides each of the tools with information about the bitstream associated with this tool.

The output data of the bitstream demultiplexer tool are:

- Depending on the type of kernel encoding the current frame, or:

- quantized and encoded noise-free spectra represented by

- information about scale factors

- arithmetically encoded spectral lines

- or linear prediction (LP) parameters together with an excitation signal represented by:

- quantized and arithmetically encoded spectral lines (coded excitation conversion, TLC), or

- ACELP coded time slot excitation

- Spectral noise filling information (optional)

- Information on the M / S solution (as an option)

- Information on the formation of temporary noise (TNS) (as an option)

- Filter bank management information

- Time Distortion Management (TW) information (optional)

- Information on managing enhanced spectral expansion through replication (repetition) of spectral bands (eSBR)

- MPEG Environment Management Information (MPEGS)

The noise-free scale factor decoding tool takes information from the bitstream demultiplexer, parses this information, and decodes the Huffman-encoded scale factors and DPCM.

The input of the decoding tool for scaling coefficients without noise is fed:

- information on the scale factors of the encoded spectra without noise.

At the output of the decoding tool for scale factors without noise appears:

- a decrypted representation in the form of integers of scale factors.

The noiseless spectrum decoding tool takes information from the bitstream demultiplexer, parses that information, decodes arithmetically encoded data, and restores the quantized spectra. The inputs to this spectrum decoding tool without noise are:

- spectra encoded without noise. The output of the spectrum decoding tool without noise appears:

- quantized values of the components of the spectra.

The inverse quantization tool takes the quantized values of the spectral components and converts the integer values to unscaled reconstructed spectra. This quantizer is a compander quantizer whose compilation coefficient depends on the selected primary coding method.

The inputs of the inverse quantization tool are:

- quantized values of the components of the spectra.

At the output of the inverse quantization tool, the following are formed:

- demagnetized inverse quantized components of the spectra.

The noise filling tool is used to fill the spectral gaps in the decoded spectra that occur when the quantized spectral values are zero, for example, due to a strong restriction on the requirement of bits in the encoder. Using a noise fill tool is optional.

The inputs of the noise filling tool are:

- demagnetized inverse quantized components of the spectra.

- Noise filling options:

- a decrypted representation in the form of integers of scale factors.

At the outputs of the noise filling tool are formed:

- unscaled inverse quantized components of the spectra that were previously quantized to zero.

- Changed representation as integers of scale factors. The rescaling tool converts the integer representation of the scale factors to the actual values and multiplies the un-scaled back-quantized spectra by the corresponding scale factors.

The inputs of the rescaling tool are:

- Decoded representation as integers of scale factors.

- Dismantled inverse quantized spectra. At the output of the rescaling tool are formed:

- Scaled back quantized spectra.

For a brief overview of the M \ S tool, please refer to ISO / IEC 14496-3, subsection 4.1.1.2.

For a quick overview of the temporary noise shaping tool (TNS), please refer to ISO / IEC 14496-3, subsection 4.1.1.2.

The filter bank / block switching tool applies the inverse frequency conversion that was performed on the encoder. The filter bank tool uses the inverse modified discrete cosine transform (IMDCT). IMDCT can be infused to provide 120, 128, 240, 256, 320, 480, 512, 576, 960, 1024, or 1152 spectral coefficients.

The filter bank is supplied with the following inputs:

- spectra (inverse quantized),

- filter bank management information.

At the output (s) of the instrument, the filter bank is formed (formed):

- Audio signal (s) recovered over a time interval.

Distorted Time Filter Bank / Switch Tool blocks replaces the usual filter bank / block switching tool when time distortion mode is acceptable. The filter bank is the same (IMDCT), which refers to a conventional filter bank, and samples of the distorted (deformed) time interval additionally processed by the window function are displayed in a linear time interval by oversampling with a time-varying frequency.

The filter bank with distorted time receives the instrument inputs:

- Inverse quantized spectra.

- Information for managing a filter bank.

- Information about managing time distortion. At the output (s) of the instrument, a filter bank with distorted time is formed (formed):

- Audio signal (s) restored on a linear time interval. Advanced SBR Tool (eSBR) restores the high frequency band of an audio signal. This is based on the replication of sequences of harmonics truncated during coding. As a result, in order to reconstruct the spectral characteristics of the original signal, a spectral envelope with a restored high-frequency band is formed, reverse filtering is applied, and noise and sinusoidal components are added. The input to the eSBR tool is:

- Quantized spectral envelope data.

- A variety of management data

- a signal on a time interval from the main AAC decoder.

The output of the eSBR tool is:

- a signal on a time interval or

- representation of the signal in the QMF region, for example, if the MPEG environment tool is used.

The MPEG Environment Tool (MPEGS) generates a plurality of signals from one or more input signals by applying a complex mixing procedure with extension to the input signal (s), which is controlled by the corresponding spatial parameters. In the context of USAC, MPEGS is used to encode a multi-channel signal by transmitting third-party parameter information along with a transmitted signal with a reduced number of channels.

The MPEGS instrument input is:

- a signal with a reduced number of channels or

- representation of the signal by the reduced number of channels in the QMF region from the eSBR tool.

The output of the MPEGS tool is formed:

- multichannel signal on a time interval.

The signal classifier tool analyzes the original input signal and generates control information from it, which causes the selection of various encoding methods. Analysis of the input signal is an implementation that depends on and tries to select the optimal coding core for a given input signal frame. The output of the signal classifier can (as an option) also be used to influence the behavior of other instruments, for example, MPEG environment, advanced SBR, filter bank with distorted time, and others.

The signal classifier is supplied to the instrument input:

- original unchanged input signal,

- additional implementation-specific parameters.

At the output of the instrument, a signal classifier is formed:

- a control signal for controlling the selection of the coding core (not LP coding of the filtered frequency domain, LP coding of the filtered frequency domain, or LP coding of the filtered time domain).

In accordance with this invention, the time / frequency resolution in block 410 in FIG. 12A and in converter 523 in FIG. 12A is controlled depending on the audio signal. The relationship between window length, conversion length, time and frequency resolution is illustrated in FIG. 13A, where it becomes clear that for a large window length, the temporal resolution decreases, but the frequency resolution becomes high, and for a short window length, the temporal resolution is high, but the frequency resolution is low .

In the first coding channel, which is preferably an AAC coding channel, the elements indicated by 410, 1201, 1202, 4021 in FIG. 12A may use various windows where the window shape is determined by a signal analyzer, which is preferably located in the signal classifier block 300, but which may also be a separate module. The encoder selects one of the windows illustrated in FIG. 13B, which have different time / frequency resolutions. The time / frequency resolution of the first long window, the second window, the fourth window, the fifth window and the sixth window is 2048 sample values for the conversion length 1024. For the short window illustrated in the third line of FIG. 13B, the time resolution is 256 sample values in accordance with window size. This corresponds to a conversion length of 128.

Similarly, the last two windows have a window length of 2304, which is better for frequency resolution than the window in the first line, but lower for time resolution. The conversion length for windows in the last two lines is 1152.

In the first coding channel, various sequences of windows that are constructed from the transform windows in FIG. 13B can be constructed. Although only a short sequence is illustrated in FIG. 13C, while other “sequences” consist of a single window, long sequences consisting of more windows can also be constructed. Note that according to FIG. 13B, for a smaller number of coefficients, that is, 960 instead of 1024, the time resolution is also lower than for the corresponding higher number of coefficients, such as 1024.

14A-14G illustrate various window resolutions / sizes in a second coding channel. In a preferred solution of the present invention, the second encoding channel has a first processing channel, which is an ACELP 526 time slot encoder, and a second processing channel including a filter bank 523. In this channel, the superframe, for example, 2048 samples in length, is divided into frames of 256 samples. Separate frames of 256 samples can be used separately so that a sequence of four windows can be applied, where each window covers two frames when using MDCT with 50 percent overlap. Then high resolution is used, as illustrated in FIG. 14D. Alternatively, when the signal allows the use of longer windows, a sequence such as that shown in FIG. 14C is used, where a double window size is applied having 1024 samples for each window (middle windows) so that one window covers four frames at 50 percent overlay.

Finally, when the signal is such that a long window can be used, this long window covers over 4096 samples again with 50 percent overlap.

In a preferred solution in which there are two channels where one channel has an ACELP encoder, the position of the ACELP frame indicated by “A” in the superframe can also determine the size of the window applied to the two adjacent TLC frames indicated by “T” in FIG. 14E. It is mainly of interest to use long windows whenever possible. However, short windows should be applied when one frame T is located between two frames A. Middle windows can be applied when there are two adjacent frames T. However, when there are three adjacent frames T, the corresponding larger window cannot be effective due to the additional complexity . Therefore, the third frame T, although it does not precede frame A, can be processed with a short window. When the whole superframe only has frames 1, then a long window can be applied.

Fig. 14F illustrates several alternatives for windows, where the window size is always 2x (twice) the number 1g of spectral coefficients due to a preferred 50 percent overlap. However, other overlap percentages can be applied to all coding channels so that the ratio between the window size and the conversion dyne can also differ from two or even close to one when no time slot matching is applied.

FIG. 14G illustrates rules for constructing a window based on the rules given in FIG. 14F. The ZL value illustrates the zeros at the beginning of the window. The value of L illustrates the number of window coefficients in the registration area. The values in part M are “1” units that do not introduce any alignment as a result of overlapping with an adjacent window that has zero values in the part corresponding to M. Part M is followed by a right overlay zone R, followed by a zone of zeros ZR that would correspond part M of the next window.

The following are materials that describe a preferred and detailed implementation of the inventive audio coding / decoding scheme, especially with respect to the decoder side.

Windows and window sequences

Quantization and coding are done in the frequency domain. To this end, a temporary. Time signal is mapped to the frequency domain in the encoder. The decoder performs the reverse mapping as described in subclause 2. Depending on the signal, the encoder can change the time / frequency resolution when using three different window sizes: 2304, 2048, and 256. To switch between the windows, the transition windows LONG_START_WINDOW, LONG_STOP_WINDOW, START_WINDOW_LPD, STOP_WINDOW_11 , STOP_START_WINDOW and STOP_START_WINDOW_1152. Table 5.11 shows the windows, the corresponding conversion length is determined and the shape of the windows is shown schematically. Three transform lengths are used: 1152, 1024 (or 960) (refer to the long transform) and 128 (or 120) coefficients (refer to the short transform).

A window sequence consists of windows so that raw_data_block always contains the data represented by 1024 (or 960) output samples. The window_sequence data element indicates the sequence of actually used windows. 13C illustrates how a sequence of windows is formed from individual windows. See subclause 2 for more information on transforms and windows.

Scaling stripes and grouping

See ISO / IEC 14496-3, clause 4, subclause 4.5.2.3.4

As described in ISO / IEC 14496-3, clause 4, subclause 4.5.2.3.4, the width of the scaling bands is based on simulating the critical bands of the human auditory system. For this reason, the number of scaling bands in the spectrum and their width depends on the conversion length and the sampling frequency. In the table. 4.110-4.128, in ISO / IEC 14496-3, clause 4, subclause 4.5.4, a list of offsets of the beginnings of each scaling band is given for transform lengths of 1024 (960) and 128 (120) and sampling frequencies. Tables originally designed for LONG_WINDOW, LONG_START_WINDOW and LONG_STOP_WINDOW are used for START_WINDOW_LPD and STOP_START_WINDOW. The offset tables for STOP_WINDOW_1152 and STOP_START_WINDOW_1152 are tab. 4-10.

Decoding Function lpd_channel_stream ()

The bit stream element lpd_channel_stream () contains all the necessary information to decode a single frame of the "linear prediction region" of the encoded signal. It receives the signal for one frame of the encoded signal, which is encoded in the LPC region, i.e. includes LPC filtering step. The residual signal of this filter (the so-called “excitation”) is then represented either by the ACELP module or in the MDCT transform region (“coded excitation transform”, TLC). To achieve a good approximation to the signal characteristics, one frame is divided into four shorter units of equal size, each of which is encoded using either the ACELP coding scheme or TLC.

This process is similar to the coding scheme described in 3GPP TS 26.290. A slightly different terminology is taken from this document, where one “superframe” denotes a signal segment of 1024 samples, where the “frame” is exactly a quarter of 1024 samples, i.e. 256 samples. Each of these frames is further subdivided into four "subframes" of equal length. Note that this terminology is used in this subsection.

Definitions, data elements

acelp_core_mode This bit field indicates the exact bit pattern in case ACELP is used as the IPD encoding mode. lpd_mode This bit field denotes the coding modes of each of the four frames in one superframe of the lpd_channel_stream () bit stream (corresponds to one AAC frame). The encoding mode is stored in the mod [] array and can take values from 0 to 3. The mapping from lpd_mode to mod [] is defined in the table below. one.

Tab. 1 - Display encoding modes for lpd_channel_stream ()

The value of the bits in the determined mode remaining mod [] elements Ipd_mode bit 4 bit 3 bit 2 bit 1 bit 0 0 ... 15 0 mod [3] mod [2] mod [1] mod [0] 16 ... 19 one 0 0 mod [3] mod [2] mod [1] = 2 mod [0] = 2 20 ... 23 one 0 one mod [1] mod [0] mod [3] = 2 mod [2] = 2 24 one one 0 0 0 mod [3] = 2 mod [2] = 2 mod [1] = 2 mod [0] = 2 25 one one 0 0 one mod [3] = 3 mod [2] = 3 mod [1] = 3 mod [0] = 3 26 ... 31 Reserved

mod [0 ... 3] The values in the mod [] array indicate the corresponding encoding modes of each frame:

Tab. 2 - Encoding modes indicated by mod []

magnitude mod [x] Frame coding mode Bitstream element 0 ACELP acelp_coding () one one frame of TCX tcx_coding () 2 TCX covering half a superframe tcx_coding () 3 TCX covering entire superframe tcx_coding ()

acelp_coding () An element of the syntactic structure that contains all the data for decoding one frame of the ACELP excitation signal. tcx_coding () A syntax structure element that contains all the data for decoding a single frame using MDCT-based coded excitation (TLC) conversion. first_tcx_flag A flag showing the currently processed TLC frame, which is the first in the superframe. lpc_data () a syntactic structure that contains all the data for decoding the set of all LPC filter parameters required to decode the current superframe. first_lpd_flag A flag indicating that the current superframe is the first in the sequence of superframes that is encoded in the LPC area. This flag can also be determined from the history of the element of the bit stream core_mode (core_mode0 and core_model in the case of channel_pair_element) in accordance with Table 3.

Tab. 3 - Definition of first_lpd_flag

core_mode - previous frame (superframe) core_mode of the current frame (superframe) first_Ipd_flag 0 one one one one 0

By analogy with [8], Section 5.2.2, there are 26 of the following ACELP or TLC options in one superframe of the lpd_channel_stream bitstream. One of these 26 mode options is indicated in the bitstream by lpd_mode. The mapping of lpd_mode to the current encoding mode for each frame in the subframe is shown in Table. 1 and table 2.

Tab. 4 - Scaling frequency bands for window length 2304 at STOP_START_1152_WINDOW and STOP_1152_WINDOW 44.1 and 48 kHz

fs [kHz] 44.1,48 num_swb_long_window 49 swb swb_offset_long_window swb swb_offset_long_window 0 0 25 216 one four 26 240 2 8 27 264 3 12 28 292 four 16 29th 320 5 twenty thirty 352 6 24 31 384 7 28 32 416 8 32 33 448 9 36 34 480 10 40 35 512 eleven 48 36 544 12 56 37 576 13 64 38 608 fourteen 72 39 640 fifteen 80 40 672 16 88 41 704 17 96 42 736 eighteen 108 43 768 19 120 44 800 twenty 132 45 832 21 144 46 864 22 160 47 896 23 176 48 928 24 196 1152

Tab. 5 - Scaling frequency bands for window length 2304 at STOP_START_1152_WINDOW and STOP_1152 WINDOW 32 kHz

fs [kHz] 32 num_swb_long_window 51 swb swb_offset_long_window swb swb_offset_long_window 0 0 26 240 one four 27 264 2 8 28 292 3 12 29th 320 four 16 thirty 352 5 twenty 31 384 6 24 32 416 7 28 33 448 8 32 34 480 9 36 35 512 10 40 36 544 eleven 48 37 576 12 56 38 608 13 64 39 640 fourteen 72 40 672 fifteen 80 41 704 16 88 42 736 17 96 43 768 eighteen 108 44 800 19 120 45 832 twenty 132 46 864 21 144 47 896 22 160 48 928 23 176 49 960 24 196 fifty 992 25 216 1152

Tab. 6 - Scaling frequency bands for window length 2304 at STOP_START_1152_WINDOW and STOP_1152_WINDOW 8 kHz

fs [kHz] 8 num_swb_long_window 40 swb swb_offset_long_window swb Swb_offset_long_window 0 0 21 288 one 12 22 308 2 24 23 328 3 36 24 348 four 48 25 372 5 60 26 396 6 72 27 420 7 84 28 448 8 96 29th 476 9 108 508

10 120 31 544 eleven 132 32 580 12 144 33 620 13 156 34 664 fourteen 172 35 712 fifteen 188 36 764 16 204 37 820 17 220 38 880 eighteen 236 39 944 19 252 1152 twenty 268

Tab. 7 - Scaling frequency bands for window length 2304 at STOP_START_1152_WINDOW and STOP_1152_WINDOW 11.025.12 and 16 kHz

fs [kHz] 11.025,12,16 num_swb_long_window 43 swb swb_offset_long_window swb swb_offset_long_window 0 0 22 228 one 8 23 244 2 16 24 260 3 24 25 280 four 32 26 300 5 40 27 320 6 48 28 344 7 56 29th 368 8 64 thirty 396 9 72 31 424 10 80 32 456 eleven 88 33 492 12 one hundred 34 532 13 112 35 572 fourteen 124 36 616 fifteen 136 37 664 16 148 38 716 17 160 39 772 eighteen 172 40 832 19 184 41 896 twenty 196 42 960 21 212 1152

Tab. 8 - Scaling frequency bands for window length 2304 at STOP_START_1152_WINDOW and STOP_1152_WINDOW 22.05 and 24 kHz

fs [kHz] 05.22 and 24 num_swb_long_window 47 swb swb_offset_long_window swb swb_offset_long_window 0 0 24 160 one four 25 172 2 8 26 188 3 12 27 204 four 16 . 28 220 5 twenty 29th 240 6 24 thirty 260 7 28 31 284 8 32 32 308 9 36 33 336 10 40 34 364 eleven 44 35 396 12 52 36 432 Ts 60 37 468 fourteen 68 38 508 fifteen 76 39 552 16 84 40 600 17 92 41 652 eighteen one hundred 42 704 19 108 43 768 twenty 116 44 832 21 124 45 896 22 136 46 960 23 148 1152

Tab. 9 - Scaling frequency bands for window length 2304 at STOP_START_1152_WINDOW and STOP 1152 WINDOW 64 kHz

fs [kHz] 64 num_swb_long_window 47 (46) swb swb_offset_long_window swb swb_offset_long_window 0 0 24 172 one four 25 192 2 8 26 216 3 12 27 240 four 16 28 268 5 twenty 29th 304 6 24 thirty 344 7 28 31 384 8 32 32 424 9 36 33 464 10 40 34 504 eleven 44 35 544 12 48 36 584 13 52 37 624 fourteen 56 38 664 fifteen 64 39 704 16 72 40 744 17 80 41 784 eighteen 88 42 824 19 one hundred 43 864 twenty 112 44 904 21 124 45 944 22 140 46 984 23 156 1152

Tab. 10 - Scaling frequency bands for window length 2304 with STOP START 1152 WINDOW and STOP 1152 WINDOW 88.2 and 96 kHz

fs [kHz] 88.2 and 96 num_swb_long_window 41 swb swb_offset_long_window swb swb_offset_long_window 0 0 21 120 one four 22 132 2 8 23 144 3 12 24 156 four 16 25 172 5 twenty 26 188 6 24 27 212 7 28 28 240 8 32 29th 276 9 36 thirty 320 10 40 31 384 eleven 44 32 448 12 48 33 512 13 52 34 576 fourteen 56 35 640 fifteen 64 36 704 16 72 37 768 17 80 38 832 eighteen 88 39 896 19 96 40 960 twenty 108 1152

Scaling Band Reference Tables

For all other references to scaling frequency bands, please refer to ISO / IEC 14496-3, clause 4, from table 4.129 of clause 4.5.4 to table 4.147.

Quantization

To quantize the spectral coefficients of AAS in the encoder, a non-uniform quantizer is used. Therefore, the decoder must invert the non-uniform quantization after decoding the Huffman scaling factors (see subclause 6.3) and decoding the data without noise (see clause 6.1).

A homogeneous quantizer is used to quantize the spectral coefficients of TLC. No inverse quantization is necessary in the decoder after decoding the spectral data without noise.

Filter Bank and Block Switching

Tool description

The time / frequency representation of the signal is displayed in the time domain and fed to the filter bank module. This module consists of an inverse modified discrete cosine transform (IMDCT), a window, and an overlay-add function. To adapt the resolution of the time / frequency of the filter bank to the characteristics of the input signal, a switching tool is used. The number N represents the length of the window, where N is the window_sequence function (see subclause 1.1). For each channel, N / 2 frequency-time values X _{i, k are} converted to x _{i, n} values on time interval N, using IMDCT. After applying the window function, for each channel, the first half z _{i, n of the} sequence is added to the second half of the previous block processed by the window function, the sequence z _{(i-1), n} , in order to restore the samples for each channel out _{i, n} to the output.

Definitions

window_sequence 2 bits indicating which window sequence is being used (i.e. block size). window_shape 1 bit showing which window function is selected.

13C shows eight window_sequences (ONLY_LONG_SEQUENCE, LONG_START_SEQUENCE, EIGHT_SHORT_SEQUENCE, LONG_STOP_SEQUENCE, STOP_START_SEQUENCE, STOP_1152_SEQUENCE, LPD_START_SEQUENCE, STOP_START_1152_SEQUENCE).

In the LPD_SEQUENCE sequence, all available window / coding mode options are referred to in the so-called codec of the linear prediction region (see section 1.3). In the context of decoding a frame encoded in the frequency domain, it is important to know whether the encoding mode of the next frame in the LP domain is indicated by LPD_SEQUENCE. However, the precise structure of the LPD_SEQUENCE allows decoding of the encoded frame in the LP region.

IMDCT Decoding Process

The analytical expression of IMDCT is:

$$x_{i,n}=\frac{2}{N}{\displaystyle \sum _{k=0}^{\frac{N}{2}-\mathrm{one}}spec\left[i\right]\left[k\right]\cos \left(\frac{2\pi }{N}\left(n+n_0\right)\left(k\frac{\mathrm{one}}{2}\right)\right)}\text{for 0}\le \text{n}<\text{N}$$

Where:

n = sample index;

i = window index;

k = spectral coefficient index;

N = window length based on window_ sequence value;

n ₀ = (N / 2 + 1) / 2.

A synthesis window of length N for the inverse transform is a function of the syntax element window_sequence and the algorithmic content. It is defined as follows:

Window Length 2304:

$$N=\{\begin{array}{l}\mathrm{2304,}\text{if STOP\_1152\_SEQUENCE}\hfill \\ \text{2304}\text{, if STOP\_START\_1152\_SEQUENCE}\hfill \end{array}$$

Window Length 2048:

$$N=\{\begin{array}{l}\mathrm{2048,}\text{if ONLY\_LONG\_SEQUENCE}\hfill \\ \mathrm{2048,}\text{if <figure-callout id="256" label="LONG\_START\_SEQUENCE" filenames="00000099.png" state="{{state}}">LONG\_START\_SEQUENCE</figure-callout>}\hfill \\ \mathrm{256,}\text{if EIGHT\_SHORT\_SEQUENCE}\hfill \\ \mathrm{2048,}\text{if LONG\_STOP\_SEQUENCE}\hfill \\ \mathrm{2048,}\text{if STOP\_START\_SEQOENCE}\hfill \\ \mathrm{2048,}\text{if LPD\_START\_SEQUENCE}\hfill \end{array}$$

The interpretations of the transition block are as follows:

$$\mathrm{and}s\text{ONLY\_LONG\_SEQUENCE in}\{\begin{array}{l}\text{ONLY\_LONG\_SEQUENCE}\hfill \\ \text{LONG\_START\_SEQUENCE}\hfill \\ \text{LPD\_START\_SEQUENCE}\hfill \end{array}$$

$$\mathrm{and}s\text{LONG\_START\_SEQUENCE in}\{\begin{array}{l}\text{EIGHT\_SHORT\_SEQUENCE}\hfill \\ \text{LONG\_STOP\_SEQUENCE}\hfill \end{array}$$

$$\mathrm{and}s\text{LONG\_STOP\_SEQUENCE in}\{\begin{array}{l}ONLY\_LONG\_SEQUENCE\hfill \\ LONG\_START\_SEQUENCE\hfill \\ LPD\_START\_SEQUENCE\hfill \end{array}$$

$$\mathrm{and}s\text{ENGHT\_SHORT\_SEQUENCE in}\{\begin{array}{l}\text{EIGHT\_SHORT\_SEQUENCE}\hfill \\ \text{LONG\_STOP\_SEQUENCE}\hfill \\ \text{STOP\_START\_SEQUENCE}\hfill \end{array}$$

$$\mathrm{and}s\text{LPD\_SEQUENCE in}\{\begin{array}{l}\text{LPD\_SEQUENCE}\hfill \\ \text{STOP\_1152\_SEQUENCE}\hfill \\ \text{STOP\_START\_1152\_SEQUENCE}\hfill \end{array}$$

$$\mathrm{and}s\text{STOP\_START\_SEQUENCE in}\{\begin{array}{l}\text{EIGHT\_SHORT\_SEQUENCE}\hfill \\ \text{LONG\_STOP\_SEQUENCE}\hfill \end{array}$$

$$\mathrm{and}s\text{LPD\_START\_SEQUENCE in}\{\text{LPD\_SEQUENCE}$$

$$\mathrm{and}s\text{STOP\_1152\_SEQUENCE in}\{\begin{array}{l}\text{ONLY\_LONG\_SEQUENCE}\hfill \\ \text{LONG\_START\_SEQUENCE}\hfill \end{array}$$

$$\mathrm{and}s\text{STOP\_START\_1152\_SEQUENCE in}\{\begin{array}{l}\text{EIGHT\_SHORT\_SEQUENCE}\hfill \\ \text{LONG\_STOP\_SEQUENCE}\hfill \end{array}$$

Window processing and block switching

Different transform windows are used depending on the window_sequence and window_shape elements. The union of the window halves describes the following representations of all possible window sequences:

For window_shape == 1, the window coefficients are determined by the Kaiser-Bessel window (KBD) as follows:

$${}^{W}KBD\_LEFT,N^{\left(n\right)}=\sqrt{\frac{{\displaystyle \sum _{p=0}^n\left[W\text{'}\text{'}\left(p,\alpha \right)\right]}}{{\displaystyle \sum _{p=0}^{N/2}\left[W\text{'}\text{'}\left(p,\alpha \right)\right]}}}\mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{<figure-callout id="2" label="N" filenames="00000074.png,00000075.png" state="{{state}}">N</figure-callout>}}{2}$$

$${}^{W}KBD\_RIGHT,N^{\left(n\right)}=\sqrt{\frac{{\displaystyle \sum _{p=0}^{N-n-\mathrm{one}}\left[W\text{'}\text{'}\left(p,\alpha \right)\right]}}{{\displaystyle \sum _{p=0}^{N/2}\left[W\text{'}\text{'}\left(p,\alpha \right)\right]}}}f\mathrm{<figure-callout\; id="2"\; label="o\; r\; N"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">o\; </figure-callout>}r\frac{\text{N}}{}\frac{}{\text{2}}\le \text{n}<N$$

Where:

W ', the Kaiser-Bessel window function, see [5], defined as:

$$W\text{'}\text{'}\left(n,\alpha \right)=\frac{I_0\left[\pi \alpha {\sqrt{10-\left(\frac{n-N/\mathrm{four}}{N/\mathrm{four}}\right)}}^2\right]}{I_0\left[\pi a\right]}dl\mathrm{I\; am}\text{0}\le \text{n}\le \frac{\text{<figure-callout id="2" label="N" filenames="00000074.png,00000075.png" state="{{state}}">N</figure-callout>}}{\text{2}}$$

$$I_0\left[x\right]={{\displaystyle \sum _{k=0}^{\infty }\left[\frac{{\left(\frac{x}{2}\right)}^k}{k!}\right]}}^2$$

α = alpha factor of the window core, $$\alpha =\{\begin{array}{l}\mathrm{four}\text{for N}=\text{2048}\left(\text{1920}\right)\hfill \\ \text{6 for N}=\text{256}\left(\text{240}\right)\hfill \end{array}$$

In another case, for window_shape == 0, a sinusoidal window is used:

$${}^{W}SIN\_LEFT,N^{\left(n\right)}=\sin \left(\frac{\pi }{N}\left(n+\frac{\mathrm{one}}{2}\right)\right)\text{for 0}\le \text{n}<\frac{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">N</figure-callout>}}{2}$$

$${}^{W}SIN\_RIGHT,N^{\left(n\right)}=\sin \left(\frac{\pi }{N}\left(n+\frac{\mathrm{one}}{2}\right)\right)\text{for}\frac{\text{<figure-callout id="2" label="N" filenames="00000074.png,00000075.png" state="{{state}}">N</figure-callout>}}{\text{2}}\le \text{n}<N$$

The window length N can be 2048 (1920) or 256 (240) for KBD and a sine window. In the case of STOP_1152_SEQUENCE and STOP_START_1152_SEQUENCE, N can be 2048 or 256, the window slopes are similar, but the flat top is longer.

Only in the case of LPD_START_SEQUENCE the right part of the window is a sinusoidal window with a length of 64 samples.

Subparagraphs a) -h) of this section show how to obtain possible window sequences.

For all types of window_sequences (window sequence), the window_shape (window shape) of the left half of the first transformation window is determined by the window shape of the previous block. The following formula expresses this fact:

$$W_{LEFT,N}\left(n\right)=\{\begin{array}{ll}{W}_{KBD\_LEFT,N}\left(n\right),\hfill & \text{if window\_shape\_previous\_block}==\text{one}\hfill \\ {\text{W}}_{\text{SIN\_LEFT}\text{, N}}\left(\text{n}\right),\hfill & \text{if window\_shape\_previous\_block}==\text{0}\hfill \end{array}$$

Where:

window_shape_previous_block (window shape of the previous block); window_shape of the previous block (i-1).

For the first decoded block first raw_data_block (), the window shape of the window_shape of the left and right halves of the window is the same.

a) ONLY_LONG_SEQUENCE:

window_sequence == ONLY_LONG_SEQUENCE The window sequence is one LONG_WINDOW with the total window length N_l equal to 2048 (1920).

When the window form is window_shape == 1, the window for ONLY_LONG_SEQUENCE is given by the following expression:

$$W\left(n\right)=\{\begin{array}{ll}{W}_{LEFT,N\_l}\left(n\right),\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_l / 2}\hfill \\ W_{KBD\_RIGHT,N\_l}\left(n\right)\hfill & for\text{N\_1 / 2}\le \text{n}<\text{N\_l}\hfill \end{array}$$

With window_shape == 0, the window for ONLY_LONG_SEQUENCE can be described by the expression:

$$W\left(n\right)=\{\begin{array}{ll}{W}_{LEFT,N\_l}\left(n\right),\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_l / 2}\hfill \\ W_{SIN\_RIGHT,N\_l}\left(n\right)\hfill & for\text{N\_1 / 2}\le \text{n}<\text{N\_l}\hfill \end{array}$$

After window processing, the values (z _{i, n} ) of the time domain can be described by the expression:

z _{i, n} = w (n) x _{i, n} ;

b) LONG_START_SEQUENCE:

A long start sequence LONG_START_SEQUENCE is needed to get the correct overlap and addition for the transition block from NLY_LONG_SEQUENCE to EIGHT_SHORT_SEQUENCE.

The window lengths N_l and N_s are set to 2048 (1920) and 256 (240), respectively.

If window_shape == 1, then the window for LONG_START_SEQUENCE is given by the following expression:

$$W\left(n\right)=\{\begin{array}{ll}{W}_{LEFT,N\_l}\left(n\right),\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_l / 2}\hfill \\ \mathrm{1.0,}\hfill & \text{for N\_l / 2}\le \text{n}<\frac{\text{3N\_l-N\_s}}{\mathrm{four}}\hfill \\ W_{KBD\_RIGHT,N\_s}\left(n+\frac{N\_s}{2}-\frac{3N\_l-N\_s}{\mathrm{four}}\right),\hfill & for\frac{\text{3N\_l-N\_s}}{\text{four}}\le n<\frac{3N\_l+N\_s}{\mathrm{four}}\hfill \\ \mathrm{0.0,}\hfill & for\frac{\text{3N\_l}+\text{N\_s}}{\text{four}}\le n<N\_l\hfill \end{array}$$

If window_shape == 0, then the window for LONG_START_SEQUENCE looks like:

$$W\left(n\right)=\{\begin{array}{ll}{W}_{LEFT,N\_l}\left(n\right),\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_l / 2}\hfill \\ \mathrm{1.0,}\hfill & \text{for N\_l / 2}\le \text{n}<\frac{\text{3N\_l-N\_s}}{\mathrm{four}}\hfill \\ W_{SIN\_RIGHT,N\_s}\left(n+\frac{N\_s}{2}-\frac{3N\_l-N\_s}{\mathrm{four}}\right),\hfill & for\frac{\text{3N\_l-N\_s}}{\text{four}}\le n<\frac{3N\_l+N\_s}{\mathrm{four}}\hfill \\ \mathrm{0.0,}\hfill & for\frac{\text{3N\_l}+\text{N\_s}}{\text{four}}\le n<N\_l\hfill \end{array}$$

The values processed by the window function can be calculated using the formula described in a).

c) EIGHT JSHORT

The window sequence window_sequence == EIGHT_SHORT consists of eight overlapping and stacked SHORT_WINDOWs (short windows) with an N_s length of 256 (240) each. The total length of the window sequence along with the previous and subsequent zeros is 2048 (1920). Each of the eight short blocks is first separately processed by the window function. The short block number is indexed by the variable j = 0, ..., M-1 (M = N_l / N_s).

The window shape of the window_shape of the previous block affects only the first of eight 'short blocks (W ₀ (n)) only. If window_shape == 1, the window function is given by:

$$W_0\left(n\right)=\{\begin{array}{ll}{W}_{LEFT,N\_s}\left(n\right),\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_s / 2}\hfill \\ W_{KBD\_RIGHT,N\_s}\left(n\right),\hfill & for\text{N\_s / 2}\le \text{n}<\text{N\_s}\hfill \end{array}$$

$$W_{\mathrm{one}-\left(M-\mathrm{one}\right)}\left(n\right)=\{\begin{array}{ll}{W}_{KBD\_LEFT,N\_s}\left(n\right)\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_s / 2}\hfill \\ W_{KBD\_RIGHT,N\_s}\left(n\right),\hfill & for\text{N\_s / 2}\le \text{n}<\text{N\_s}\hfill \end{array}$$

In another case, window_shape == 0, the window function can be described as:

$$W_0\left(n\right)=\{\begin{array}{ll}{W}_{LEFT,N\_s}\left(n\right),\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_s / 2}\hfill \\ W_{KBD\_RIGHT,N\_s}\left(n\right),\hfill & for\text{N\_s / 2}\le \text{n}<\text{N\_s}\hfill \end{array}$$

$$W_{\mathrm{one}-\left(M-\mathrm{one}\right)}\left(n\right)=\{\begin{array}{ll}{W}_{SIN\_LEFT,N\_s}\left(n\right),\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{N\_s / 2}\hfill \\ W_{SIN\_RIGHT,N\_s}\left(n\right),\hfill & for\text{N\_s / 2}\le \text{n}<\text{N\_s}\hfill \end{array}$$

Overlapping and adding up a sequence of eight short windows EIGHTJ3HORT window_sequence is obtained by processing the window function of the values z _{i, n} in the time domain and is described by the expression:

$$z_{i,n}=\{\begin{array}{ll}0\hfill & for\text{0}\le \text{n}<\frac{\text{N\_l-n\_s}}{\text{four}}\hfill \\ {\text{x}}_{\text{0}\text{, n-}\frac{\text{N\_l-n\_s}}{\text{four}}}\cdot W_0\left(n-\frac{N\_l-N\_s}{\mathrm{four}}\right),\hfill & for\frac{\text{N\_l-n\_s}}{\mathrm{four}}\le n<\frac{N\_l+N\_s}{\mathrm{four}}\hfill \\ x_{j-\mathrm{one,}n-\frac{N\_l+\left(2j-3\right)\cdot N\_s}{\mathrm{four}}}\cdot W_{j-\mathrm{one}}\left(n-\frac{N\_l+\left(2j-3\right)N\_s}{\mathrm{four}}\right)+x_{j,n-\frac{N\_l+\left(2j-\mathrm{one}\right)N\_s}{\mathrm{four}}}\cdot W_j\left(n-\frac{N\_l+\left(2j-\mathrm{one}\right)N\_s}{\mathrm{four}}\right),\hfill & for\text{one}\le \text{j}<\text{M}\text{,}\frac{\text{N\_l}+\left(\text{2j-1}\right)N\_s}{\text{four}}\le n<\frac{N\_l+\left(2j+\mathrm{one}\right)N\_s}{\mathrm{four}}\hfill \\ x_{M-\mathrm{one,}n-\frac{N\_l+\left(2M-3\right)N\_s}{\mathrm{four}}}\cdot W_{M-\mathrm{one}}\left(n-\frac{N\_l+\left(2M-3\right)N\_s}{\mathrm{four}}\right),\hfill & for\frac{\text{N\_l}+\left(\text{2M-1}\right)N\_s}{\mathrm{four}}\le n<\frac{N\_l+\left(2M+\mathrm{one}\right)N\_s}{\mathrm{four}}\hfill \\ 0\hfill & for\frac{\text{N\_l}+\left(\text{2M}+\text{one}\right)N\_s}{\mathrm{four}}\le n<N\_l\hfill \end{array}$$

d) LONG_STOP_SEQUENCE

This window sequence is necessary to switch from EIGHT_SHORT_SEQUENCE back to ONLY_LONG_SEQUENCE.

If window_shape == 1 window for LONG_STOP_SEQUENCE is given by the following expression:

$$W\left(n\right)=\{\begin{array}{ll}\mathrm{0.0,}\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{N\_l-N\_s}{\mathrm{four}}\hfill \\ W_{LEFT,N\_s}\left(n-\frac{N\_l-N\_s}{\mathrm{four}}\right),\hfill & \text{for}\frac{\text{N\_l-n\_s}}{\mathrm{four}}\le n<\frac{N\_l+N\_s}{\mathrm{four}}\hfill \\ \mathrm{1.0,}\hfill & for\frac{\text{N\_l}+\text{N\_s}}{\text{four}}\le n<N\_l/2\hfill \\ W_{KBD\_RIGHT,N\_l}\left(n\right),\hfill & for\text{N\_l / 2}\le \text{N\_l}\hfill \end{array}$$

If window_shape == 0 the window for LONG_START_SEQUENCE is defined by:

$$W\left(n\right)=\{\begin{array}{ll}\mathrm{0.0,}\hfill & \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{N\_l-N\_s}{\mathrm{four}}\hfill \\ W_{LEFT,N\_s}\left(n-\frac{N\_l-N\_s}{\mathrm{four}}\right),\hfill & \text{for}\frac{\text{N\_l-n\_s}}{\mathrm{four}}\le n<\frac{N\_l+N\_s}{\mathrm{four}}\hfill \\ \mathrm{1.0,}\hfill & for\frac{\text{N\_l}+\text{N\_s}}{\text{four}}\le n<N\_l/2\hfill \\ W_{SIN\_RIGHT,N\_l}\left(n\right),\hfill & for\text{N\_l / 2}\le \text{N\_l}\hfill \end{array}$$

The values processed by the window function over a time interval can be calculated using the formula described in a).

e) STOP_START_SEQUENCE:

The STOP_START_SEQUENCE element is needed to get the correct overlap and addition for the transition block from EIGHT_SHORT_SEQUENCE to EIGHT_SHORT_SEQUENCE, when only a long ONLY_LONG_SEQUENCE sequence is needed.

The window lengths N_l N_s are 2048 (1920) and 256 (240), respectively.

If window_shape == 1, the window for STOP_START_SEQUENCE is given by the following expression:

$$W\left(n\right)=\{\begin{array}{cc}\mathrm{0.0,}& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{N\_l-n\_s}}{\mathrm{four}}\\ W_{LEFT,N\_s}\left(n-\frac{N\_l-N\_s}{\mathrm{four}}\right),& for\frac{\text{N\_l-n\_s}}{\text{four}}\le n<\frac{N\_l+N\_s}{\mathrm{four}}\\ \mathrm{1.0,}& for\frac{\text{N\_l}+\text{N\_s}}{\text{four}}\le n<\frac{3N\_l-N\_s}{\mathrm{four}}\\ W_{KBD\_RIGHT,N\_s}\left(n+\frac{N\_s}{2}-\frac{3N\_l-N\_s}{\mathrm{four}}\right),& for\frac{\text{3N\_l-N\_s}}{\text{four}}\le n<\frac{3N\_l+N\_s}{\mathrm{four}}\\ \mathrm{0.0,}& for\frac{\text{3N\_l}+\text{N\_s}}{\text{four}}\le n<N\_l\end{array}$$

If window_shape == 0, the window for STOP_START_SEQUENCE looks like:

$$W\left(n\right)=\{\begin{array}{cc}\mathrm{0.0,}& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{N\_l-n\_s}}{\mathrm{four}}\\ W_{LEFT,N\_s}\left(n-\frac{N\_l-N\_s}{\mathrm{four}}\right),& for\frac{\text{N\_l-n\_s}}{\text{four}}\le n<\frac{N\_l+N\_s}{\mathrm{four}}\\ \mathrm{1.0,}& for\frac{\text{N\_l}+\text{N\_s}}{\text{four}}\le n<\frac{3N\_l-N\_s}{\mathrm{four}}\\ W_{SIN\_RIGHT,N\_s}\left(n+\frac{N\_s}{2}-\frac{3N\_l-N\_s}{\mathrm{four}}\right),& for\frac{\text{3N\_l-N\_s}}{\text{four}}\le n<\frac{3N\_l+N\_s}{\mathrm{four}}\\ \mathrm{0.0,}& for\frac{\text{3N\_l}+\text{N\_s}}{\text{four}}\le n<N\_l\end{array}$$

The values processed by the window function over a time interval can be calculated using the formula described in a).

f) LPD_START_SEQUENCE:

The LPD_START_SEQUENCE element is needed to get the correct overlap and addition for the transition block from ONLY_LONG_SEQUENCE to LPD_SEQUENCE.

The window lengths N_l and N_s are 2048 (1920) and 256 (240), respectively.

If window_shape == 1, the window for LPD_START_SEQUENCE is given by:

$$W\left(n\right)=\{\begin{array}{cc}{W}_{LEFT,N\_l}\left(n\right),& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{<figure-callout id="2" label="N\_l" filenames="00000074.png,00000075.png" state="{{state}}">N\_l</figure-callout>}}{2}\\ \mathrm{1.0,}& f\mathrm{<figure-callout\; id="2"\; label="o\; r\; N\_l"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">o\; </figure-callout>}r\frac{\text{N\_l}}{}\frac{}{\text{2}}\le n<\frac{3N\_l-N\_s}{\mathrm{four}}\\ W_{KBD\_RIGHT,N\_\raisebox{1ex}{$s$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}\left(n+\frac{N\_s}{\mathrm{four}}-\frac{3N\_l-N\_s}{\mathrm{four}}\right),& for\frac{\text{3N\_l-N\_s}}{\text{four}}\le n<\frac{3N\_l}{\mathrm{four}}\\ \mathrm{0.0,}& for\frac{\text{3N\_l}}{\text{four}}\le n<N\_l\end{array}$$

If window_shape == 0, the window for LPD_START_SEQUENCE looks like:

$$W\left(n\right)=\{\begin{array}{cc}{W}_{LEFT,N\_l}\left(n\right),& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{<figure-callout id="2" label="N\_l" filenames="00000074.png,00000075.png" state="{{state}}">N\_l</figure-callout>}}{2}\\ \mathrm{1.0,}& f\mathrm{<figure-callout\; id="2"\; label="o\; r\; N\_l"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">o\; </figure-callout>}r\frac{\text{N\_l}}{}\frac{}{\text{2}}\le n<\frac{3N\_l-N\_s}{\mathrm{four}}\\ W_{SIN\_RIGHT,N\_\raisebox{1ex}{$s$}\!\left/ \!\raisebox{-1ex}{$2$}\right.}\left(n+\frac{N\_s}{\mathrm{four}}-\frac{3N\_l-N\_s}{\mathrm{four}}\right),& for\frac{\text{3N\_l-N\_s}}{\text{four}}\le n<\frac{3N\_l}{\mathrm{four}}\\ \mathrm{0.0,}& for\frac{\text{3N\_l}}{\text{four}}\le n<N\_l\end{array}$$

The values processed by the window function over a time interval can be calculated using the formula described in a).

g) STOP_1152_SEQUENCE:

The STOP_1152JSEQUENCE element is necessary to get the correct overlap and addition for the transition block from LPDJSEQUENCE to ONLY_LONG_SEQUENCE.

The window lengths N_l and N_s are 2048 (1920) and 256 (240), respectively.

If window_shape == 1, the window for STOP_1152_SEQUENCE is given by:

$$W\left(n\right)=\{\begin{array}{cc}\mathrm{0.0,}& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{N\_l}}{\mathrm{four}}\\ W_{LEFT,N\_s}\left(n-\frac{N\_l}{\mathrm{four}}\right),& for\frac{\text{N\_l}}{\text{four}}\le n<\frac{N\_l+2N\_s}{\mathrm{four}}\\ \mathrm{1.0,}& for\frac{\text{N\_l}+2\text{N\_s}}{\text{four}}\le n<\frac{2N\_l+3N\_s}{\mathrm{four}}\\ W_{KBD\_RIGHT,N\_l}\left(n+\frac{N\_l}{2}-\frac{2N\_l+3N\_s}{\mathrm{four}}\right),& for\frac{\text{2N\_l-3N\_s}}{\text{four}}\le n<N\_l+\frac{3N\_s}{\mathrm{four}}\\ \mathrm{0.0,}& for\text{N\_l}+\frac{\text{3N\_s}}{\text{four}}\le n<N\_l+N\_s\end{array}$$

If window_shape === 0, the window for STOP_1152JSEQUENCE looks like:

$$W\left(n\right)=\{\begin{array}{cc}\mathrm{0.0,}& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{N\_l}}{\mathrm{four}}\\ W_{LEFT,N\_s}\left(n-\frac{N\_l}{\mathrm{four}}\right),& for\frac{\text{N\_l}}{\text{four}}\le n<\frac{N\_l+2N\_s}{\mathrm{four}}\\ \mathrm{1.0,}& for\frac{\text{N\_l}+2\text{N\_s}}{\text{four}}\le n<\frac{2N\_l+3N\_s}{\mathrm{four}}\\ W_{SIN\_RIGHT,N\_l}\left(n+\frac{N\_l}{2}-\frac{2N\_l+3N\_s}{\mathrm{four}}\right),& for\frac{\text{2N\_l}+\text{3N\_s}}{\text{four}}\le n<N\_l+\frac{3N\_s}{\mathrm{four}}\\ \mathrm{0.0,}& for\text{N\_l}+\frac{\text{3N\_s}}{\text{four}}\le n<N\_l+N\_s\end{array}$$

The values processed by the window function over a time interval can be calculated using the formula described in a).

h) STOP_START_1152_SEQUENCE:

The STOPJ3TART_1152_SEQUENCE element is needed to get the correct overlap and addition for the transition block from LPD_SEQUENCE to EIGHT_SHORT_SEQUENCE when only the long ONLY_LONG_SEQUENCE sequence is needed.

The window lengths N_l and N_s are 2048 (1920) and 256 (240), respectively. If window_shape == 1, the window for STOP_START_SEQUENCE is given by:

$$W\left(n\right)=\{\begin{array}{cc}\mathrm{0.0,}& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{N\_l}}{\mathrm{four}}\\ W_{LEFT,N\_s}\left(n-\frac{N\_l}{\mathrm{four}}\right),& for\frac{\text{N\_l}}{\text{four}}\le n<\frac{N\_l+2N\_s}{\mathrm{four}}\\ \mathrm{1.0,}& for\frac{\text{N\_l}+2\text{N\_s}}{\text{four}}\le n<\frac{3N\_l}{\mathrm{four}}+\frac{N\_s}{2}\\ W_{KBD\_RIGHT,N\_s}\left(n+\frac{N\_s}{2}-\frac{3N\_l}{\mathrm{four}}+\frac{N\_s}{2}\right),& for\frac{\text{3N\_l}}{\text{four}}+\frac{N\_s}{2}\le n<\frac{3N\_l}{\mathrm{four}}+N\_s\\ \mathrm{0.0,}& for\frac{\text{3N\_l}}{\text{four}}+N\_s\le n<N\_l+N\_s\end{array}$$

If window_shape == 0, the window for STOP_START_SEQUENCE looks like:

$$W\left(n\right)=\{\begin{array}{cc}\mathrm{0.0,}& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{N\_l}}{\mathrm{four}}\\ W_{LEFT,N\_s}\left(n-\frac{N\_l}{\mathrm{four}}\right),& for\frac{\text{N\_l}}{\text{four}}\le n<\frac{N\_l+2N\_s}{\mathrm{four}}\\ \mathrm{1.0,}& for\frac{\text{N\_l}+2\text{N\_s}}{\text{four}}\le n<\frac{3N\_l}{\mathrm{four}}+\frac{N\_s}{2}\\ W_{SIN\_RIGHT,N\_s}\left(n+\frac{N\_s}{2}-\frac{3N\_l}{\mathrm{four}}+\frac{N\_s}{2}\right),& for\frac{\text{3N\_l}}{\text{four}}+\frac{N\_s}{2}\le n<\frac{3N\_l}{\mathrm{four}}+N\_s\\ \mathrm{0.0,}& for\frac{\text{3N\_l}}{\text{four}}+N\_s\le n<N\_l+N\_s\end{array}$$

The values processed by the window function over a time interval can be calculated using the formula described in a).

Overlap and add with previous window sequence

In addition to overlapping and adding up a sequence of eight short EIGHT_SHORT window_sequence windows, the first (left) part of each window_sequence window sequence overlaps and adds to the second (right) part of the previous window_sequence window sequence, which leads to the final value of the values out _{i, n} in the time domain.

A mathematical description of this operation can be given as follows.

In the case of ONLY_LONG_SEQUENCE, LONG_START_SEQUENCE, EIGHT_SHORT_SEQUENCE, LONG_STOP_SEQUENCE, STOP_START_SEQUENCE, LPD_START_SEQUENCE:

; $$for\text{ 0}\le \text{n}<\frac{\text{N}}{\text{2}}$$

, N=2048 (1920) $$ou{t}_{i,n}=z_{i,n}+z_{i-\mathrm{one,}n+\frac{\mathrm{<figure-callout\; id="2"\; label="N"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">N</figure-callout>}}{2}}$$

; $$\mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{<figure-callout id="2" label="N" filenames="00000074.png,00000075.png" state="{{state}}">N</figure-callout>}}{\text{2}}$$

, N = 2048 (1920)

And in the case of STOP_1152_SEQUENCE, STOP_START_1152_SEQUENCE:

; $$for\text{ 0}\le \text{n}<\frac{\text{N\_l}}{\text{2}}$$

, N_l=2048, N_s=256 $$ou{t}_{i,n}=z_{i,n}+z_{i-\mathrm{one,}n+\frac{N\_\mathrm{<figure-callout\; id="2"\; label="l"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">l</figure-callout>}}{2}+\frac{3N\_s}{\mathrm{four}}}$$

; $$\mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{<figure-callout id="2" label="N\_l" filenames="00000074.png,00000075.png" state="{{state}}">N\_l</figure-callout>}}{\text{2}}$$

, N_l = 2048, N_s = 256

In the case of LPD_START_SEQUENCE, the next sequence is LPDJSEQUENCE. To get good overlap and addition, the SIN or KBD window is applied to the left side of LPD_SEQUENCE.

$$for\text{ 0}\le \text{n}<\frac{\text{N}}{\text{2}}$$

With N=128 $$W_{SIN\_LEFT,N}\left(n\right)=\sin \left(\frac{\pi }{N}\left(n+\frac{\mathrm{one}}{2}\right)\right)$$

$$\mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\frac{\text{<figure-callout id="2" label="N" filenames="00000074.png,00000075.png" state="{{state}}">N</figure-callout>}}{\text{2}}$$

With N = 128

In the case of STOP_1152_SEQUENCE, STOP_START_1152_SEQUENCE, the previous sequence is LPD_SEQUENCE. To get good overlap and addition, the TDAC window is applied to the left of LPD_SEQUENCE.

IMDCT

See subclause 2.3.1

Window function processing and block switching

Depending on the window_shape element, various prototypes of conversion windows with an excessive sampling frequency are used; the length of windows with an excessive sampling frequency is defined as

N _OS = 2 n_long os_factor_win

For window_shape == 1, window coefficients are given by the Kaiser-Bessel window (KBD) as follows:

$$W_{KBD}\left(n-\frac{N_{OS}}{2}\right)=\sqrt{\frac{{\displaystyle \sum _{p=0}^{N_{OS-n-\mathrm{one}}}\left[W\left(p,\alpha \right)\right]}}{{\displaystyle \sum _{p=0}^{N_{OS}/2}\left[W\left(p,\alpha \right)\right]}}}fo\mathrm{<figure-callout\; id="2"\; label="r\; N\; OS"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">r\; </figure-callout>}\frac{{\text{N}}_{\text{OS}}}{}\frac{{}_{}}{2}\le n<N_{OS}$$

Where:

W, the Kaiser-Bessel window function, see [5], defined as:

$$W\text{'}\text{'}\left(n,\alpha \right)=\frac{\left[\pi \alpha \sqrt{1.0-\left(\frac{n-N_{OS}/\mathrm{four}}{N_{OS}/\mathrm{four}}\right)}\right]}{I_0\left[\pi \alpha \right]}\mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}\le \frac{{\text{<figure-callout id="2" label="N OS" filenames="00000074.png,00000075.png" state="{{state}}">N </figure-callout>}}_{\text{OS}}}{2}$$

$$I_0\left[x\right]={\displaystyle \sum _{k=0}^{\infty }{\left[\frac{{\left(\frac{x}{2}\right)}^k}{k!}\right]}^2}$$

α = alpha factor of the window core, α = 4

On the other hand, for window-shape == 0, a sinusoidal window (SIN) is used:

$$W_{SIN}\left(n-\frac{N_{OS}}{2}\right)=\sin \left(\frac{\pi }{N_{OS}}\left(n+\frac{\mathrm{one}}{2}\right)\right)fo\mathrm{<figure-callout\; id="2"\; label="r\; N\; OS"\; filenames="00000074.png,00000075.png"\; state="\{\{state\}\}">r\; </figure-callout>}\frac{{\text{N}}_{\text{OS}}}{}\frac{{}_{}}{2}\le n<N_{OS}$$

For all types of window_sequences, the prototype used for the left side of the window is determined by the window shape of the previous block. The following formula expresses this fact:

$$left\_window\_shape\left[n\right]=\{\begin{array}{l}{W}_{KBD}\left[n\right],\text{if window\_shape\_previous\_block}==\text{one}\hfill \\ {\text{W}}_{\text{Sin}}\left[n\right],\text{if window\_shape\_previous\_block}==\text{0}\hfill \end{array}$$

Similarly, the prototype for the correct form of the window is defined by the following formula:

$$right\_window\_shape\left[n\right]=\{\begin{array}{l}{W}_{KBD}\left[n\right],\text{if window\_shape}==\text{one}\hfill \\ {\text{W}}_{\text{Sin}}\left[n\right],\text{if window\_shape}==\text{0}\hfill \end{array}$$

Since transition lengths are already defined, this should only be differentiated between EIGHT_SHORT_SEQUENCE and all others:

a) EIGHT SHORT SEQUENCE:

The following C code describes window function processing and internal overlap-addition of the EIGHT_SHORT_SEQUENCE sequence:

MDCT based TCX

Tool description

When core_mode is 1, and when one or more of the three TCX coding modes is selected as the "linear prediction region", that is, one of the 4 elements of the mod [] array is greater than 0, the MDCT-based TCX tool is used. An MDCT-based TCX receives quantized spectral coefficients from an arithmetic decoder. Before applying the inverse MDCT transform, the quantized coefficients are first supplemented with comfort noise to obtain a weighted synthesized signal in the time domain, which is then fed to the LPC filter of the weighted synthesized signal.

Definitions

Ig The number of quantized spectral coefficients at the output of an arithmetic decoder noise_factor Quantization Noise Level Index noise level Reconstructed Spectrum Noise Level noise [] Noise generated vector global_gain Resampling Quantization Gain Index g rescaling gain rms The average square of the synthesized signal in the time domain x [], x [] Synthesized time-domain signal

Decoding process

MDCT-based TLC receives from the arithmetic decoder a plurality of Ig quantized spectral coefficients, which are determined by the mod [] and last_lpd_mode values. These two values also determine the length and shape of the window to be applied in the reverse MDCT. The window is made up of three parts, the left side of the sample overlap L, the middle part of the samples M and the right side of the sample overlap R. To get a 2 * log length MDCT window, ZL zeros and ZR zeros on the right are added on the left, as shown in Fig. 14G for tab. 3 / Fig. 14F.

Tab. 3 - Number of spectral coefficients as a function of last_lpd_mode and mod []

Last_Ipd_mode values Value mod [x] The number of Ig spectral coefficients Zl L M R Zr 0 one 320 160 0 256 128 96 0 2 576 288 0 512 128 224 0 3 1152 512 128 1024 128 512 1 ... 3 one 256 64 128 128 128 64 1 ... 3 2 512 192 128 384 128 192 1..3 3 1024 448 128 896 128 448

The MDCT window is given by

$$W\left(n\right)=\{\begin{array}{cc}0& \mathrm{<figure-callout\; id="0"\; label="f\; o\; r"\; filenames="00000093.png,00000099.png"\; state="\{\{state\}\}">f\; </figure-callout>}or\text{0}\le \text{n}<\text{Zl}\\ W_{SIN\_LEFT,L}\left(n-ZL\right)& \text{for zl}\le \text{nZL}+\text{L}\\ \mathrm{one}& \text{for zl}+\text{L}\le \text{n}<\text{Zl}+\text{L}+\text{M}\\ W_{SIN\_RIGHT,R}\left(n-ZL-L-M\right)& \text{for zl}+\text{L}+\text{M}\le \text{Zl}+\text{L}+\text{M}+\text{<figure-callout id="0" label="R" filenames="00000093.png,00000099.png" state="{{state}}">R</figure-callout>}\\ 0& \text{for zl}+\text{L}+\text{M}+\text{R}\le \text{n}<\text{2lg}\end{array}$$

Quantized spectral coefficients quant [], formed by an arithmetic decoder and added by comfortable noise. The input noise level is determined by the decoded parameter noise_factor as follows:

noise_level = 0.0625 * (8-noise_factor)

Then, using the random function random_sign (), generating random variables -1 or +1, the noise vector noise [] is calculated.

noise [i] = random_sign () * noise_level;

The vectors quant [] and noise [] combine to form the vector of reconstructed spectral coefficients r [], in a way in which 8 consecutive zeros in quant [] are replaced with noise components noise []. Sequences of 8 non-zero values are determined in accordance with the formula:

$$\{\begin{array}{l}rl\left[i\right]=\mathrm{one}fori\in \left[0.1g/6\right]\hfill \\ rl\left[\lg /6+i\right]={\displaystyle \sum _{k=0}^7|\; |\; |quant\left[\lg /6+8.\left[i/8\right]+k\right]|\; |\; |fori\in \left[\lg /6\right]}\hfill \end{array}$$

Then the reconstructed spectrum is obtained as:

$$r\left[i\right]=\{\begin{array}{l}quant\left[i\right]ifrl\left[i\right]=\mathrm{one}\hfill \\ noise\left[i\right]otherwise\hfill \end{array}$$

Before using inverse MDCT, spectral smoothing is applied in accordance with the following steps:

1. Calculation of the energy E _{m of an} 8-dimensional block with index m for each 8-dimensional block of the first quarter of the spectrum.

2. Calculation of the expression R _m = sqrt (E _m / E _I ), where I is the block index with the maximum value of all E _m .

3. if R _m <0.1, then R _m = 0.1

4. if R _m <R _m-1 , then R _m = R _m-1

Each 8-dimensional block lying in the first quarter of the spectrum is then multiplied by the parameter R _m .

The reconstructed (reconstructed) spectrum is fed to the inverse MDCT transform. The output signal x [] processed by the window function is rescaled using the gain parameter g obtained by inverting the quantization of the decoded global_gain index:

g = 10 ^{global_gain / 28 / (2.rms)}

Where rms is calculated as:

$$rms=\sqrt{\frac{{\displaystyle \sum _{i=\lg /2}^{3*\lg /2-\mathrm{one}}{x}^2\left[i\right]}}{L+M+R}}$$

Then the demapped signal synthesized in the time domain is equal to:

x _w [i] = x [i] · g

After de-scaling, window function processing and overlap / addition are applied.

The reconstructed TLC signal x (n) is then filtered using the filter A (z) (1-αz ^-1 ) / (A (z / λ) to find the excitation signal that will go to the synthesis filter. Note what kind of subframe to filter an interpolating LP filter is used. Once the excitation is determined, the signal is restored by passing the excitation through a 1 / Â (z) synthesis filter and then through a 1 / (1-0.68z ^-1 ) filter, as described above.

Note that excitation is also necessary in order to update the ACELP adaptive code table and allow switching from TLC to ACELP in a subsequent frame. We also note that the duration of TLC synthesis is given by the length of the TLC frame (without overlapping): 256.512 or 1024 samples for mod [] values of 1.2 or 3, respectively.

Definitions

Definitions can be found in ISO / IEC 14496-3, subsection 1, subclause 1.3 (Terms and definitions) and in 3GPP TS 26.290, section 3 (Definitions and abbreviations).

Although some aspects have been described in the context of a device, it is clear that these aspects also represent a description of the corresponding method, where the unit or device corresponds to a method step or a feature of a method step. Likewise, aspects described in the context of a method step also provide a description of the corresponding block or item, or features of the corresponding device.

The audio signal encoded according to the invention may be stored on a digital storage medium or may be transmitted on a transmission medium, such as a wireless transmission medium or a wired transmission medium, such as the Internet.

Depending on certain requirements, embodiments of the invention may be implemented in hardware or in software. The embodiment can be performed using a digital storage medium, for example, a diskette, DVD, CD, ROM, EPROM, EEPROM or FLASH memory, on which electronically readable control signals are stored that can be used by a programmable computer system, so that the corresponding method will be executed.

Some embodiments of the invention include a storage medium onto which readable control signals are electronically recorded that can be used by a programmable computer system such that one of the methods described herein is performed.

In general, embodiments of the present invention can be implemented as a computer program product with program code serving to execute one of the methods when the computer program is executed on a computer. The program code may, for example, be stored on a computer-readable medium.

Other embodiments include a computer program in order to execute one of the methods described herein stored on a computer-readable medium.

In other words, an embodiment of the invented method is then a computer program having program code for executing one of the methods described herein when the computer program is executed on a computer.

A further embodiment of the invented methods is then a storage medium (either a digital storage medium or a computer readable medium) comprising a computer program in order to execute one of the methods described herein.

A further embodiment of the invented method is then a data stream or a sequence of signals representing a computer program in order to execute one of the methods described here. A data stream or a sequence of signals may, for example, be configured to be transmitted through a data transmission system, for example, via the Internet.

A further embodiment includes a processing means, for example, a computer, or a programmable logic device, configured to implement one of the methods described herein.

A further embodiment includes a computer on which a computer program is then installed in order to perform one of the methods described herein.

In some embodiments, a programmable logic device (eg, a programmable logic integrated circuit) may be used, configured to implement some or all of the methods described herein. In some embodiments, a programmable logic integrated circuit may interact with a microprocessor to perform one of the methods described herein. In general, the methods are preferably performed by any hardware-based device.

The embodiments described above are merely illustrations for embodying the principles of the present invention. It is understood that modifications and changes to the quantities and details described herein will be apparent to those skilled in the art. Therefore, it is intended to be limited only by the claims and not by certain details presented by way of description and explanation of embodiments.

Literature

[1] ISO / IEC 11172-3: 1993, Information technology - Coding of moving pictures and associated audio for digital storage media at up to about 1,5 Mbit / s, Part 3: Audio.

[2] ITU-T Rec. H.222.0 (1995) ISO / IEC 13818-1: 2000, Information technology - Generic coding of moving pictures and associated audio information: - Part 1: Systems.

[3] ISO / IEC 13818-3: 1998, Information technology - Generic coding of moving pictures and associated audio information: - Part 3: Audio.

[4] ISO / IEC 13818-7: 2004, Information technology - Generic coding of moving pictures and associated audio information: - Part 7: Advanced Audio Coding (AAC).

[5] WHEELS 14496-3: 2005, Information technology - Coding of audio-visual objects - Part 1: Systems

[6] ISO / IEC 14496-3: 2005, Information technology - Coding of audio-visual objects - Part 3: Audio

[7] ISOAEC 23003-1: 2007, Information technology - MPEG audio technologies - Part 1: MPEG Surround

[8] 3GPP TS 26.290 V6.3.0, Extended Adaptive Multi-Rate - Wideband (AMR-WB +) codec; Transcoding functions

[9] 3GPP TS 26.190, Adaptive Multi-Rate - Wideband (AMR-WB) speech codec; Transcoding functions

[10] 3GPP TS 26.090, Adaptive Multi-Rate (AMR) speech codec; Transcoding functions

Claims (21)

1. An audio encoder for encoding an audio signal including a first encoding channel (400) in order to encode an audio signal using a first encoding algorithm to obtain a first encoded audio signal, wherein the first encoding channel comprises a first time / frequency converter (410) for converting the input audio signal into a spectral region;
a second encoding channel (500) in order to encode the audio signal using the second encoding algorithm to obtain a second encoded audio signal, where the first encoding algorithm is different from the second encoding algorithm, and the second encoding channel includes a region converter in order to convert the input audio a signal from an input region to an audio signal of an output region, and a second converter (523) in order to convert the input audio signal to a spectral region;
a switch (200) for switching between the first encoding channel and the second encoding channel so that for a part of the input audio signal, either the first encoded audio signal or the second encoded audio signal is in the audio output of the encoder;
an audio signal analyzer (300, 525) in order to analyze a portion of the audio signal to determine whether a portion of the audio signal is represented as a first encoded audio signal or a second encoded audio signal in the audio output of the encoder, where the audio signal analyzer is further configured to determine an appropriate non-constant time / frequency resolution of the first converter and the second converter when the first encoded audio signal or the second encoded audio signal is generated, Representing a part of the audio signal; and
an output interface (800) for generating an output audio signal of an encoder including a first encoded audio signal and a second encoded audio signal and information indicative of a first encoded audio signal and a second encoded audio signal and information indicative of time / frequency resolution used to encode the first encoded audio signal and to encode the second encoded audio signal. 2. The audio encoding device according to claim 1, wherein the audio signal analyzer (300, 525) is configured to classify a portion of the audio signal as a speech-like audio signal or a music-like audio signal, and is configured to detect a transient in the case of a music audio signal for in order to determine the time / frequency resolution of the first time / frequency converter (410) or in order to perform analysis analysis through synthesis in order to determine the time / frequency resolution of the second converter (523). 3. The audio encoder according to claim 1, wherein the first time / frequency converter (410) and the second converter (523) include a variable window function conversion processor including a variable window size window function and a variable length conversion function, and where the audio signal analyzer (300/525) is configured to control based on the analysis of the audio signal, window size and / or conversion length. 4. The audio encoding device according to claim 1, in which the second encoding channel includes a first processing channel (522), configured to process the audio signal in the region defined by the region converter (510), and a second processing channel (523, 524), including the second converter (523), where the audio signal analyzer is configured to divide a portion of the audio signal into a sequence of sub-parts, and where the audio signal analyzer is configured to determine the time / frequency resolution of the second converter (523) depending dix subpart treated in the first treatment channel relative to subpart portion processed in the second processing channel. 5. The audio encoder according to claim 4, wherein the first processing channel includes an ACELP encoder (526), in which the second processing channel includes MDCT-TCX processing units (527), in which the audio signal analyzer (300/525) is configured the ability to set a high value of the temporal resolution of the second converter, determined by the length of the sub-part or a relatively low value of the time resolution, determined by the length of the sub-part, multiplied by an integer value greater than one. 6. The audio encoding device according to claim 1, in which the audio signal analyzer (300, 525) is configured to determine the classification of the audio signal in a constant raster covering a plurality of blocks of audio samples of the same size, and is configured to divide the block into a variable number of blocks depending from an audio signal, where the length of the sub-block determines in the first case the time / frequency resolution or in the second case the frequency resolution. 7. The audio encoding device according to claim 1, wherein the audio signal analyzer (300, 525) is configured to determine a time / frequency resolution to select from at least two different lengths of the set of different window lengths from the set of 2304, 2048, 256, 1920, 2160, 240 samples, or choose from different transform lengths different transform lengths, including at least two of the set consisting of 1152, 1024, 1080, 960, 128, 120 coefficients of the transform block, or in which the analyzer the audio signal (300, 525) is configured to the time / frequency resolution of the second converter as one of many different window lengths, many different window lengths, which are at least two of 640, 1152, 2304, 512, 1024 or 2048 samples, or as one of many different conversion lengths, various transform lengths, including at least two of the set consisting of 320, 576, 1152, 256, 512, 1024 spectral coefficients of the transform block. 8. The audio encoder according to claim 1, wherein the second encoding channel includes a first processing channel (522) in order to process the audio signal;
a second processing channel including a second converter; and
further a switch (521) in order to switch between the first processing channel (522) and the second processing channel (523, 524) so that for a part of the input audio signal into the second encoding channel or the first processed audio signal or the second processed audio signal second encoded audio signal. 9. A method of audio encoding an audio signal, including encoding an audio signal in a first encoding channel (400) using a first encoding algorithm to obtain a first encoded audio signal, a first encoding channel including a first time / frequency converter (410) in order to convert input audio signal in the spectral region;
encoding an audio signal in a second encoding channel (500) using a second encoding algorithm to obtain a second encoded audio signal, where the first encoding algorithm is different from the second encoding algorithm, a second encoding channel including a region converter in order to convert the input audio signal from the input region into the output region, and the second Converter (523) in order to convert the input audio signal into a spectral region;
switching (200) between the first coding channel and the second coding channel so that for a part of the input audio signal, either the first encoded audio signal or the second encoded audio signal of the encoder in the audio output signal;
analyzing (300, 525) the portion of the audio signal to determine whether the portion of the audio signal is represented as a first encoded audio signal or a second encoded audio signal in the audio output of the encoder,
determining an appropriate non-constant time / frequency resolution of the first converter and the second converter when the first encoded audio signal or the second encoded audio signal representing a part of the generated audio signal is generated; and
generating (800) an output audio signal of an encoder including a first encoded audio signal and a second encoded audio signal and information indicative of a first encoded audio signal and a second encoded audio signal, and information indicative of a time / frequency resolution used to encode the first encoded audio signal and in order to encode a second encoded audio signal. 10. An audio decoder for decoding an encoded audio signal including a first encoded audio signal, a second encoded audio signal, a flag indicating a first encoded audio signal and a second encoded audio signal, and time / frequency resolution information to be used for in order to decode the first encoded audio signal and the second encoded audio signal, the audio decoder includes
the first decoding channel (431, 440) in order to decode the first encoded audio signal using the first controllable frequency / time converter (440), the first controllable frequency / time converter configured to control using time / frequency resolution information for the first an encoded audio signal to obtain a first decoded audio signal;
a second decoding channel in order to decode the second encoded audio signal using a second controlled frequency / time converter (534), a second controlled frequency / time converter (534) configured to control using time / frequency resolution information for the second encoded audio signal;
a controller (990) for controlling the first frequency / time converter (440) and the second frequency / time converter (534) using time / frequency resolution information;
a region converter (540) in order to generate a synthesized audio signal using a second decoded audio signal; and
combiner (604) in order to combine the first decoded audio signal and the synthesized audio signal to obtain a decoded audio signal. 11. The audio decoder of claim 10, in which the controller (990) is configured to control the first frequency / time converter (440) and the second converter the frequency / time converter (534) so that for the first frequency / time converter (440) resolution time / frequency was selected from many different window lengths, different window lengths, which are at least two of 2304, 2048, 256, 1920, 2160, 240 samples, or selected from many different conversion lengths, various conversion lengths, including at least two of the group consisting of and 1152, 1024, 1080, 960, 128, 120 coefficients of the conversion unit, or for the second frequency / time converter (534), the time / frequency resolution was selected as one of many different window lengths, many different window lengths, which are at least two from 640, 1152, 2304, 512, 1024 or 2048 samples, or was chosen from a variety of different conversion lengths, various conversion lengths, including at least two from the group consisting of 320, 576, 1152, 256, 512, 1024 spectral conversion block coefficients. 12. The audio decoder of claim 10, wherein the second decoding channel includes a first reverse processing channel (531) for reverse processing the first processed audio signal, which is further included in the encoded audio signal to obtain a first back-processed audio signal;
where the second controlled frequency / time converter (534) is located in the second reverse processing channel, configured to invert the processing of the second encoded audio signal in an area identical to the region of the first back-processed audio signal to obtain a second back-processed audio signal;
a further combiner (532) in order to combine the first back-processed audio signal and the second back-processed audio signal to obtain a combined audio signal; and
where the combined audio signal enters the combiner (600). 13. The audio decoder of claim 10, where the first frequency / time converter (440) and the second frequency / time converter are time-alignment converters having an overlap / addition unit (440 s) in order to cancel time-alignment areas included in the first encoded audio signal and in the second encoded audio signal. 14. The audio decoder of claim 10, wherein the encoded audio signal includes an encoded identification of method information, whether the encoded audio signal is a first encoded audio signal or a second encoded audio signal, and
where the decoder further includes an input interface (900) in order to interpret the encoded method information to determine whether the encoded audio signal should be supplied to the first decoding channel or to the second decoding channel. 15. The audio decoder according to claim 1, wherein the first encoded audio signal is arithmetically encoded and where the first encoding channel includes an arithmetic decoder. 16. The audio decoder according to claim 1, in which the first coding channel includes a decanter having a non-uniform de-quantization characteristic in order to cancel the result of the non-uniform quantization applied in generating the first encoded audio signal, where the second coding channel includes a de-quantizer using other characteristics dequantization, or where the second coding channel does not include the dequantizer. 17. The audio decoder according to claim 1, in which the controller (990) is configured to control the first frequency / time converter and the second frequency / time converter, applying for each converter a discrete frequency / time resolution from among various different discrete frequency / time resolutions, the number of possible different frequency / time resolutions, which is large for the second converter, compared to the number of possible different frequency / time resolutions for the first converter. 18. The audio decoder of claim 10, wherein the region converter is an LPC synthesis processor (544) that generates a synthesized audio signal using LPC filter information, LPC filter information included in the encoded audio signal. 19. A method for audio decoding an encoded audio signal, an encoded audio signal including a first encoded audio signal, a second encoded audio signal, a flag indicative of a first encoded audio signal and a second encoded audio signal, and time / frequency resolution information to be used for to decode the first encoded audio signal and the second encoded audio signal, including decoding the first decoding channel (431, 440) of the first encoded audio signal using the first controlled frequency / time converter (440), the first controlled frequency / time converter configured to control using time / frequency resolution information for the first encoded audio signal to obtain a first decoded audio signal;
decoding a second decoding channel of a second encoded audio signal using a second controlled frequency / time converter (534), a second controlled frequency / time converter (534) configured to control using time / frequency resolution information for the second encoded audio signal;
controlling (990) a first frequency / time converter (440) and a second frequency / time converter (534) using time / frequency resolution information;
generating (540) a converter of a region of synthesized audio signal using a second decoded audio signal; and
combining (604) a first decoded audio signal and a synthesized audio signal to obtain a decoded audio signal. 20. A computer-readable medium comprising a program recorded thereon, which causes the computer processor to carry out the steps of the method of claim 9. 21. A computer-readable medium comprising a program recorded thereon, which causes the computer processor to carry out the steps of the method of claim 19.

Images