RU2344493C2 - Sound coding with different durations of coding frame - Google Patents

Abstract

FIELD: information technologies.

SUBSTANCE: invention is related to method of sound signal coding support, in which at least one segment of sound signal should be coded with the help of coding model, which makes it possible to use different durations of coding frame, according to which it is suggested to define at least one control parameter on the basis of sound signal characteristics. Then this control parameter is used for limitation of versions of possible frame durations selection in respect to at least one segment of signal. Group of inventions also comprises module (10, 11), in which this method is realised, device (1) and system, which comprise such module (10, 11), and also software product, which includes program code for realisation of suggested method.

EFFECT: presentation of possibility of simple selection of corresponding most suitable duration of coding frame.

34 cl, 4 dwg

Description

FIELD OF TECHNOLOGY

The invention relates to a method for supporting encoding an audio signal, in which at least one segment of the specified audio signal must be encoded using an encoding model that allows the use of different encoding frame durations. The invention also relates to a corresponding module, to a corresponding electronic device, to a corresponding system, and to a corresponding software product.

BACKGROUND

It is known that the encoding of audio signals makes it possible to ensure their efficient transmission and / or storage.

The audio signal may be a speech signal or another type of audio signal, for example music, and different coding models may be suitable for different types of audio signals.

A widespread technique for encoding speech signals is coding according to a linear prediction algorithm with excitation by an algebraic code (ACELP - Algebraic Code-Exited Linear Prediction). ACELP models a human speech system and is well suited for encoding the frequency of a speech signal. As a result, high speech quality can be achieved at very low transmission rates. For example, the Adaptive Multi-Rate Wideband (AMR-WB) codec is a speech codec that is based on ACELP technology. The AMR-WB codec was described, for example, in 3GPP TS 26.190 technical documentation: "Speech Codec speech processing functions; AMR Wideband speech codec; Transcoding functions" (AMR wideband speech codec; transcoding functions), V5.1.0 (2001-12). However, for other types of audio signals, such as music, speech codecs, which are based on a person’s speech education system, usually work quite poorly.

A widely used method of encoding audio signals other than speech is transformation coding (TLC - transform coding). The benefits of transform coding for an audio signal are based on the masking effect of perception and coding in the frequency domain. The quality of the resulting audio signal can be further improved by choosing a suitable encoding frame length for transform coding. But at the same time, that conversion coding methods lead to high quality of sound signals other than speech, they have poor performance for periodic speech signals. Therefore, the quality of speech encoded with the conversion is usually quite low, especially with large TLC frame lengths.

The advanced AMR-WB codec (AMR-WB +) encodes a stereo audio signal as a high-speed mono signal and provides some additional information regarding stereo extension. The AMR-WB + codec uses both coding models, ACELP and TLC, to encode the main monaural signal in the frequency band from 0 to 6400 Hz. For the TLC model, an encoding frame duration of 20, 40, or 80 ms is used.

Since the ACELP model can reduce sound quality, and conversion coding usually does not work well for speech, especially when long frames are used, you must choose the best coding model accordingly. The choice of the coding model that actually needs to be applied can be done in a variety of ways.

In systems requiring the use of algorithms of low complexity, such as mobile multimedia services (MMS), music / speech classification algorithms are usually used to select the optimal encoding model. These algorithms classify the entire source signal either as music or as speech based on a study of the energy and frequency of the audio signal.

If an audio signal consists only of speech or only music, then using the same coding model, which is based on such a classification of music / speech for the whole signal, will be satisfactory. However, in many other cases, the audio signal to be encoded is a mixed type audio signal. For example, speech may be present in an audio signal along with music at the same time and / or alternate with it.

In these cases, the separation of entire source signals into a music or speech category is too limited. In this case, only switching between coding models during coding of the audio signal can maximize the overall sound quality. That is, the ACELP model is also partially used for encoding an original signal classified as an audio signal other than a speech signal, and the TLC model is also partially used for an original signal classified as a speech signal.

The advanced AMR-WB codec (AMR-WB +) is also designed to encode such mixed types of audio signals using heterogeneous encoding models that use the principle of frame division.

The selection of coding models in AMR-WB + can be done in several ways.

In the most complex approach, the signal is first encoded using all possible combinations of ACELP and TLC models. Further, for each combination, the signal is synthesized again. Then, based on the quality of the synthesized speech signals, the best excitation is selected. The quality of the synthesized speech signal obtained for a particular combination can be measured, for example, by determining its signal-to-noise ratio (SNR). This synthesis analysis method provides good results. However, in some applications, due to its very high complexity, it cannot be put into practice. To a large extent, the complexity is caused by ACELP encoding, which is the most complex part of the encoder.

For example, in systems such as MMS, the full closed loop synthesis analysis method is too complicated. Therefore, in the MMS encoder, to determine which coding model — ACELP or TLC — to choose for encoding a single frame, an open-loop method is used, which has low complexity

The AMR-WB + codec for choosing the appropriate encoding model for each frame offers two different methods of small complexity with an open loop. Both open-loop methods for selecting the appropriate encoding model evaluate the characteristics of the original signal and the encoding parameters.

In the first open-loop method, the sound signal within each frame is first divided into several frequency bands and the relationship between the energy in the lower and upper frequency bands, as well as the change in the energy level in these bands, is examined. Then, based on both measurements taken or on different combinations of these measurements using different analysis windows and decision threshold values, the audio content in each frame of the audio signal is classified as music-like or speech-like.

In the second open-loop method, which is also called refinement of the classification of the model, the choice of coding model is based on an assessment of the periodicity and stationarity characteristics of the audio content in the corresponding frame of the audio signal. The characteristics of periodicity and stationarity are more accurately estimated on the basis of determining the correlation, the parameters of the long term prediction (LTP - Long Term Prediction) and measuring the spectral distance.

If the signal properties are examined using the open-loop method to select either ACELP or TLC, and TLC is selected for encoding, then it is still necessary to choose one of the TLC frame durations of 20, 40 or 80 ms, which should be used. However, the optimal frame duration for TLC is very difficult to choose based on the characteristics of the signal in the open-loop method.

Thus, it is possible to select TLC frame durations only within the framework of the above analysis through synthesis. However, for systems requiring the use of algorithms of low complexity, the method of analysis through synthesis is too complicated, even if it is used only to select TLC frame durations.

SUMMARY OF THE INVENTION

The aim of the invention is to enable effective and simple selection of the length of the encoding frame, which must be used to encode the length of the audio signal.

A method is proposed for supporting coding of an audio signal, in which at least one segment of the indicated audio signal must be encoded using an encoding model that allows the use of different encoding frame durations. The proposed method includes determining at least one control parameter based at least in part on the characteristics of said audio signal. In addition, the proposed method includes limiting the selection of possible durations of the coding frame for at least one segment by means of the specified at least one control parameter.

In addition, an audio coding support module is proposed in which at least one section of said audio signal needs to be encoded using an encoding model that allows the use of different encoding frame durations. The module comprises a parameter selection unit capable of determining at least one control parameter based at least in part on the characteristics of the audio signal. In addition, the module comprises a block for selecting a frame duration, which is configured to limit the selection of possible encoding frame durations for at least one segment of the audio signal by means of at least one control parameter provided by the first estimation unit. This module may be, for example, an encoder or part of an encoder.

In addition, an electronic device is proposed that includes such a module.

In addition, the proposed system for encoding sound, which contains such a module and additionally a decoder for decoding audio signals that have been encoded with variable lengths of the encoding frame.

Finally, a software product is proposed in which the code for the audio encoding support program is stored. At least one segment of the audio signal must be encoded using a coding model that allows the use of different coding frame durations. In the process of execution in the processing part of the encoder, the program code performs the steps of the proposed method.

The invention proceeds from the fact that although the final determination of the length of the coding frame for a particular segment of the audio signal often cannot be performed based on the characteristics of the signal, nevertheless, such characteristics of the signal allow you to pre-select the appropriate duration of the coding frame. Therefore, it is proposed to determine at least one control parameter based on the characteristics of the signal for the corresponding segment of the audio signal and use this at least one parameter to limit the available options for choosing the encoding frame duration.

An advantage of the invention is that it reduces the number of options for choosing a coding frame duration using an approach having low complexity. On the other hand, decreasing the encoding frame length selection options reduces the complexity of the final encoding frame length selection to be used.

In one embodiment of the invention, the final selection of the length of the coding frame is performed by synthesis analysis. That is, in the event that after the proposed restriction there is more than one choice of possible durations of the encoding frame, then each of the remaining durations of the encoding frame with conversion is used to encode at least one segment. Then, the resulting encoded signals are decoded again with the corresponding used transform coding frame length. Now you can select the encoding frame duration, which leads to the best decoded audio signal in at least one segment.

Due to the preceding limitation, it is possible to significantly reduce the number of required analysis cycles through synthesis in comparison with the aforementioned full closed loop method. As a result, the overall complexity of the encoder in which the invention is implemented is also reduced.

The best decoded audio signal can be determined in various ways. For example, it can be determined by comparing signal-to-noise ratios obtained using each of the remaining coding frame durations. Signal-to-noise ratios can be easily determined and provide a reliable indication of signal quality.

In the event that several encoding models can be used to encode an audio signal, for example, the TLC model and the ACELP model, it is also necessary to determine which encoding model for which segment of the audio signal should be used. This can be achieved with a low complexity method based on the characteristics of the audio signal for the corresponding segment, as mentioned above. Then, the number and / or position of the segments for which a different encoding model should be used than that which allows using different encoding frame durations can also be used as a control parameter to limit the options for choosing the encoding frame duration.

For example, the length of the encoding frame cannot exceed the size of the segment or segments between two segments for which a different encoding model was chosen.

In further embodiments of the invention, the length of the coding frame is selected only within the corresponding super-segment, which contains a predetermined number of segments. In this case, the options for choosing the encoding frame duration for a particular segment can also be limited based on the boundaries of the super-segment to which the segment belongs.

For example, such a super-segment can be a super-frame that contains four frames of an audio signal as segments, each frame of an audio signal has a duration of 20 ms. If the encoding model is a TLC model, then it can provide the encoding frame durations of 20, 40, or 80 ms. If in this case, for example, for the second frame of the audio signal in the superframe, the ACELP encoding model was selected, it is known that the third frame of the audio signal can be encoded with a coding duration of no more than 20 ms or together with the fourth frame of the audio signal 40 ms.

In another preferred embodiment of the invention, an indicator showing whether to apply a larger or shorter encoding frame duration provides an additional control parameter. An indication that a shorter encoding frame length should be used, at the same time eliminating at least the option of choosing the longest coding frame duration, and an indication that a longer coding frame length should be used excludes at least the option of choosing the shortest coding frame duration .

BRIEF DESCRIPTION OF THE DRAWINGS

Other objects and features of the present invention will become apparent from the following detailed description, together with the accompanying drawings.

Figure 1 is a diagram of a system for encoding sound in accordance with an embodiment of the invention.

FIG. 2 is a flowchart illustrating an embodiment of a method in accordance with the invention implemented in the system of FIG. 1;

Figure 3 is a first table illustrating a restriction imposed on combinations of operating modes, which is based on control parameters in accordance with the invention; and

4 is a second table illustrating a restriction imposed on combinations of operating modes that is based on control parameters in accordance with the invention.

DETAILED DESCRIPTION OF THE INVENTION

Figure 1 is a diagram of a system for encoding sound in accordance with an embodiment of the invention, which allows you to select the duration of the encoding frame for the transform coding model.

The system comprises a first device 1 including an AMR-WB + encoder 10, and a second device 2 including an AMR-WB + decoder 20. The first device 1 may be, for example, an MMS server, and the second device 2 may be, for example, a mobile phone.

The first device 1 comprises a first evaluation unit 12 for a first choice of an open-loop encoding model. In addition, the first device 1 comprises a second evaluation unit 13 for refining the first selection by an additional open-loop method and in parallel for determining the short frame indicator as one control parameter. The first evaluation unit 12 and the second evaluation unit 13 together form a parameter selection unit. In addition, the first device 1 comprises a TLC frame duration selection unit 14 for limiting the options for selecting the encoding frame duration in the event that a TLC model is selected, and for choosing the best option among the remaining options with an open loop method. In addition, the first device 1 comprises an encoding unit 15. The encoding unit 15 is capable of applying the ACELP encoding model, the TLCX encoding model, which uses a TLC frame length of 20 ms, the TLC40 encoding model, which uses a TLC frame length of 40 ms, or the TLC80 encoding model, which uses a TLC frame length of 80 ms, to apply to received sound frames.

The first evaluation unit 12 is connected to the second evaluation unit 13 and to the encoding unit 15. The second evaluation unit 13 is furthermore connected to the TLC frame duration selecting unit 14 and to the encoding unit 15. The TLC frame duration selection unit 14 is also connected to the encoding unit 15.

You must understand that the presented blocks 12-15 are intended for encoding a monophonic audio signal, which can be formed from a stereo audio signal. Additional stereo information may be generated in additional stereo units not shown. In addition, it should be noted that the encoder 10 contains additional blocks not shown. In addition, it must be understood that the presented blocks 12-15 do not have to be separate blocks, but can be equally mixed with each other or with other blocks.

Blocks 12, 13, 14 and 15 can be implemented, in particular, using SW software operating in the processing part 11 of the encoder 10, which is shown by a dashed line.

Next, the processing in the encoder 10 will be described in more detail with reference to the algorithm shown in figure 2.

Processing is performed for the corresponding superframe. Each superframe has a duration of 80 ms and contains four consecutive frames of the audio signal.

The encoder 10 receives the audio signal supplied by the first device 1. The audio signal is converted into a monophonic audio signal and a linear prediction filter (LP) and computes linear prediction coding in each frame in order to simulate the spectral envelope.

The first estimation unit 12, in the first open-loop analysis, processes the resulting LPC excitation produced by the LP filter for each frame in the superframe. This analysis, based on the characteristics of the original signal, determines whether the contents of the corresponding frame can be considered as a speech signal or other audio content similar to music. The analysis, as mentioned above, can be based, for example, on the estimation of energy in different frequency bands. For each frame that supposedly contains a speech signal, an ACELP encoding model is selected, and for each frame that supposedly contains a different audio signal, a TLC model is selected. At this point in time, there is no separation between TLC models that use different encoding frame lengths. For those frames for which the characteristics under study do not clearly indicate the content with a speech signal or the content with a different type of audio signal, an undefined state is selected.

The first evaluation unit 12 informs the encoding unit 15 of all frames for which the ACELP model has so far been selected.

Then, the second evaluation unit 13 performs a second open-loop analysis using a single-frame principle of operation for further splitting into ACELP and TLC frames, which is based on the characteristics of the signal. In parallel, the second evaluation unit 13 determines the indicator of the short frame - the MoMtcx flag as one control parameter. If the MoMtcx flag is set, then the use of TCX80 is prohibited.

The processing in block 13 of the second rating for the corresponding frame is performed only if the indicator of voice activity is set for it - the VAD flag and if block 12 of the first rating did not select the ACELP coding model for this frame.

If the result of the calculation of the first open-loop analysis using node 12 of the first estimate is an undefined state, then the spectral distance is calculated first and a lot of available signal characteristics are typed.

The spectral distance SD _{n of the} current frame n is calculated from the parameters of an Immittance Spectral Pair (ISP) according to the following equation.

where ISP _n is the vector of coefficients of the ISP frame n and where ISP _n (i) is the i-th element of this vector. The ISP parameters are available in one way or another, since the LP coefficients are converted to the ISP domain for quantization and interpolation purposes.

The Lag _n parameter contains the values of the two open-cycle delays of the current frame. Lag is a long-term filter delay. Usually it is equal to the true period of the fundamental tone or a multiple of it, or it is a multiple of it. An open-loop pitch analysis is performed twice per frame, that is, every 10 ms, to find two estimates of the pitch delay in each frame. This is done in order to simplify the analysis of the fundamental tone and limit the search for the fundamental tone in a closed loop to a small number of delays near estimates of delays in an open loop.

Next, LagDif _buf is a buffer containing the open cycle delay values of the previous ten frames lasting 20 ms.

Gain _n contains two LTP gain values for the current frame n.

The NormCorr _n parameter contains two normalized correlation values for the current frame n.

The MaxEnergy _buf parameter is the maximum value of the buffer containing the energy values. The energy buffer contains the energy values of the current frame n and the previous five frames, each lasting 20 ms.

Now encoding modes are selected, and the NoMtcx control parameter is set in accordance with the following open-loop algorithm:

if (SD _n > 0.2)

Mode = ACELP_MODE;

else

if (LagDif _buf <2)

if (Lag _n == HIGH LIMIT or Lag _n == LOW LIMIT) {

if (Gainn-NormCorr _n <0.1 and NormCorr _n > 0.9)

Mode = ACELP_MODE

else

Mode = TCX_MODE

else if (Gainn-NormCorr _n <0.1 and NormCorr _n > 0.88)

Mode == ACELP_MODE

else if (Gain _n -NormCorr _n > 0.2)

Mode == TCX_MODE

else

NoMtcx = NoMtcx + 1

if (MaxEnergybuf <60)

if (SD _n > 0.15)

Mode = ACELP_MODE;

else NoMtcx = NoMtcx + 1.

Thus, the various characteristics of the signal and their combinations are compared with various predefined threshold values to determine if the frame of the indefinite mode contains speech or other audio content, and to assign a suitable encoding model. Similarly, depending on some of these characteristics of the signal and their combinations, a short frame indicator is set - the NoMtcx flag.

If the result of the calculations of the first open-loop analysis using block 12 of the first evaluation was TLC mode, then, on the contrary, it is determined whether the VAD flag was reset to zero for at least one frame in the previous superframe. If so, then the indicator of a short frame - the NoMtcx flag is also set to '1'.

If the encoding mode for the current frame is already set to TLC mode or is still set to indefinite mode, the decision about the mode is checked further. For this purpose, the spectral envelope vector mag subjected to the discrete Fourier transform (DFT) is first created from the LP coefficients of the current frame. Then, the encoding mode is checked in accordance with the following algorithm:

if (Gain _n -NormCorr _n <0.006 and NormCorr _n > 0.92 and Lag _n > 21)

DFTSum = 0;

for (i = l; i <40; i ++) {

DFTSum = DFTSum + mag [i];

if (DFTSum> 95 and mag [0] <5) {

Mode = TCX_MODE;

else

Mode = ACELP_MODE;

NoMtcx = NoMtcx + 1

Thus, the final sum of DFTSum is the sum of the first 40 elements of the mag vector, excluding its first element mag (0).

The second evaluation unit 13 additionally informs the encoding unit 15 of all frames for which the ACELP model has been selected.

In TLC frame duration selection section 14, control parameters are first evaluated to limit the number of TLC frame duration selection options.

One control parameter is the number of ACELP modes selected in the superframe. In the event that the ACELP encoding model was selected for four frames in a superframe, there is not a single frame left for which the TLC frame duration needs to be determined. In the event that the ACELP encoding model was selected for three frames in a superframe, then the TLC frame duration is set to 20 ms.

Further restrictions are introduced based on the table of FIGS. 3 or 4. FIGS. 3 and 4 show corresponding tables of five columns linking the TLC frame durations that can be selected with various combinations of selected coding modes.

Both tables in the first column show seven possible combinations of the selected encoding modes for four frames in a superframe. In each combination, no more than two ACELP modes are selected. The combinations are (0,1,1,1), (1,0,1,1), (1,1,0,1), (1,1,1,0), (1,1,0,0 ), (0,0,1,1) and (1,1,1,1), the last combination occurs twice. In this representation of the selected combinations, '0' stands for ACELP mode, and '1' stands for TLC mode.

The corresponding fourth column presents the Aind control parameter, which for each combination in the first column indicates the number of selected ACELP operating modes. You can see that there are only combinations of modes that correspond to the values Aind '0', '1' and '2', since if the values are equal to '3' or '4', the TLC frame duration selection unit 14 can immediately select the duration TLC frame without further processing.

The corresponding fifth column presents the indicator of a short frame - the NoMtcx flag. This parameter is evaluated by the TLC frame duration selection unit 14 only if the Aind control parameter has a value of '0', that is, if the ACELP operation mode has not been selected for any frame in the superframe.

The corresponding second and third columns for each combination show the TLC frame durations that can be selected for TLC mode frames, taking into account the restrictions imposed by the control parameters. For each combination in the first column, check for no more than two TLC frame durations. In these combinations of TLC frame durations, '0' means an ACELP coding frame with a duration of 20 ms, '1' means a TLC frame with a duration of 20 ms, a sequence of two '2' is a TLC frame with a duration of 40 ms and a sequence of four '3' is a TLC frame with a duration of 80 ms .

For example, for the first combination of operating modes (0,1,1,1), combinations of coding frame durations (0,1,1,1) and (0,1,2,2) are allowed. That is, either the second, third and fourth frames are encoded as a TLC frame with a duration of 20 ms or the second frame is encoded as a TLC frame with a duration of 20 ms, and the third and fourth frames are encoded as a TLC frame with a duration of 40 ms.

Similarly, for the second combination of operating modes (1,0,1,1), a combination of coding frame durations (1,0,1,1) and (1,0,2,2) is allowed. For the third combination of operating modes (1,1,0,1), combinations of coding frame durations (1,1,0,1) and (2,2,0,1) are allowed, For the fourth combination of operating modes (1,1,1 , 0) combinations of encoding frame durations (1,1,1,0) and (2,2,1,0) are allowed. For the fifth combination of operating modes (1,1,0,0), a combination of coding frame durations (1,1,0,0) and (2,2,0,0) is allowed. For the sixth combination of operating modes (0,0,1,1), a combination of coding frame durations (0,0,1,1) and (0,0,2,2) is allowed.

For the seventh combination of operating modes (1,1,1,1), the short frame indicator - the NoMtcx flag indicates whether to try longer or shorter TCX frame duration. The NoMtcx flag is set for a superframe in the event that the second evaluation unit 13 has set it for at least one of the frames in the superframe. If the NoMtcx flag is set for a superframe, then only short frame durations are allowed.

In the table of FIG. 3, this means that the TLC frame duration selection section 14 immediately selects a TLC frame duration of 20 ms for the whole superframe. That is, the only allowed combination of TLC frame durations is (1,1,1,1). In the table of Fig. 4, the set NoMtcx flag means that a combination of TLC frame durations (1,1,1,1) and, additionally, a combination of TLC frame durations (2,2,2,2) are allowed, the latter means a TLC frame duration of 40 ms.

If the indicator of a short frame is the NoMtcx flag is not set, then only large TLC frame durations are allowed. In the table in FIGS. 3 and 4, this means that combinations of TLC frame durations (2,2,2,2) and (3,3,3,3) are allowed, the latter means a single TLC frame with a duration of 80 ms.

Optimal encoding of clean music usually requires longer TLC frames, and speech is obviously best encoded using ACELP. Especially at the beginning of a musical and / or speech signal, when the energy is low or the voice activity indicator VAD was reset to zero in previous frames, longer TLC frames used to encode speech impair its quality. On the other hand, short TLC frames of 20 ms duration are comparatively good for music and certain speech segments. With some characteristics of the signal, it is difficult to determine whether the contents of the frame are music or speech. Therefore, in this case, a short TLC frame is a good alternative to the optimal coding model because it is suitable for both types of content. Thus, the short frame indicator is well suited as a control parameter.

Additional combinations of coding frame durations for the presented combinations of operating modes are not allowed by the encoder structure, in which the TCX40 model is not allowed for the central frames of the audio signal.

Similarly, additional combinations of operating modes at Aind <3, not shown in FIGS. 3 and 4, allow only a single combination of coding frame durations, either on their own or because of the encoder structure. That is, a combination of modes (1,0,0,1) allows only a combination of coding frame durations (1,0,0,1), and a combination of modes (0,1,1,0) allows only a combination of coding frame durations (0, 1,1,0).

Since the control parameters Aind and NoMtcx limit the combination of modes with regard to the TLC frame durations, for each superframe it is necessary to check no more than two frame durations.

In the event that there are two possible combinations of TLC frame durations, then in order to find the optimal TLC model or models for the superframe, an algorithm such as signal-to-noise ratio comparison is used in block TLC frame duration selection 14.

To estimate the TLC frame durations that can be selected, the frames in the superframe for which TLC mode has been selected are encoded using transform coding and with both allowed combinations of TLC frame duration. For example, the TLC model is based on the Fast Fourier Transform (FFT). The encoded signals are decoded again, and then the results for both TLC frame durations are compared based on a segmented signal-to-noise ratio.

The signal-to-noise ratio is the signal-to-noise ratio for one subframe of the TLC frame. The subframe has a length N, which corresponds to a subframe of the original audio signal with a duration of 5 ms.

The signal to noise segment ratio in subframe i (segSNR _i ) for each subframe of the TLC frame is determined according to the following equation:

- это амплитуда кодированного и затем декодированного звукового сигнала в позиции n в подкадре.In this equation, x _w (n) is the amplitude of the digitized source audio signal at position n in the subframe, a

is the amplitude of the encoded and then decoded audio signal at position n in the subframe.

Based on this, the average segment signal-to-noise ratio is determined for all subframes in the TLC frame according to the following equation:

where N _SF is the number of subframes in the TLC frame. Since the TLC frame may have a duration of 20, 40, or 80 ms, N _SF may be 4, 8, or 16.

Then, the TLC frame duration selection section 14 determines which of the allowed TLC frame durations for a certain number of frames of the audio signal leads to a better average signal to noise ratio. For example, if each of the two frames of the audio signal could be encoded using the TCX20 model or together using the TCX40 model, then the average signal-to-noise ratio of the TCX40 frame is compared with the average sum of the signal-to-noise ratios for both TCX20 frames. The TLC frame duration, which leads to a higher average signal-to-noise ratio, is selected and reported to the coding unit 15.

The encoding unit 15 encodes all frames of the audio signal using an appropriately selected encoding model indicated by either the first evaluation unit 12, the second evaluation unit 13, or the TLC frame duration selecting unit 14. For example, TLC is based on FFT using the selected coding frame duration, and ACELP coding uses, for example, LTP and fixed codebook parameters to drive the LPC.

Then, the encoding unit 15 provides encoded frames for transmission to the second device 2. In the second device 2, the decoder 20 decodes all received frames using the ACELP encoding model or using one of the TLC models. Decoded frames are provided, for example, for playback to the user of the second device 2.

Thus, the presented choice of TLC frame duration is based on a semi-closed cycle approach in which the main type of coding model and control parameters are selected using the open loop method, and then the TLC frame duration is selected from a limited number of options using the closed loop method. While in the analysis process with a full closed loop, analysis through synthesis is always performed four times per superframe, in this semi-closed loop method, analysis through synthesis needs to be performed no more than two times per superframe.

It should be noted that the described embodiment of the invention represents only one of the many possible options for its implementation.

Claims (34)

1. A method of supporting encoding an audio signal, in which at least one segment of the specified audio signal must be encoded using the encoding model, which allows you to use different durations of the encoding frame, the specified method comprises:
determining at least one control parameter at least partially based on the characteristics of said audio signal; and
the limitation of these options for choosing the possible duration of the coding frame for the specified at least one segment by means of the specified at least one control parameter. 2. The method according to claim 1, which comprises determining said at least one control parameter based on at least one of the following parameters:
an indicator of the spectral distance between the current and previous frames;
the number of frames in the superframe that are selected for encoding using another encoding model. 3. The method according to claim 1, which contains:
in the event that after the specified restriction there remains more than one option for choosing the possible durations of the encoding frame, encoding the specified at least one segment using each of the indicated remaining durations of the encoding frame with conversion;
decoding said encoded segments with the corresponding used encoding frame length with conversion and
selection for the specified at least one segment of such a length of the encoding frame, which leads to the best decoded audio signal on the specified at least one segment. 4. The method according to claim 3, in which the duration of the encoding frame, which leads to a better decoded segment, is determined by comparing the signal-to-noise ratio obtained for each of the indicated encoding frame durations. 5. The method according to claim 4, in which for the specified signal-to-noise ratio of the audio signal obtained at a particular encoding frame duration, first, the signal-to-noise segment ratio is determined separately for a plurality of subframes in the corresponding encoding frame, and then said signal-to-noise segment ratios are specified the subframes of the encoding frame are averaged over the entire encoding frame to obtain the specified signal to noise ratio for the specified at least one segment. 6. The method according to claim 1, in which for each segment of the specified audio signal is determined based on the characteristics of the audio signal for the corresponding segment, whether to apply the specified coding model or another coding model, and the specified at least one control parameter contains an indication of the segments, for which the specified other coding model has been selected. 7. The method of claim 6, wherein said encoding model is a transform encoding model, and said other encoding model is a linear prediction algorithm with an algebraic code excitation algorithm. 8. The method according to claim 6, in which each segment of the specified audio signal has a predetermined duration, and the indicated indication of the segments for which another encoding model was selected is provided for the corresponding super-cut, which contains a predetermined number of these segments. 9. The method according to claim 1, in which each segment of the specified audio signal has a predetermined duration, while a predetermined number of consecutive segments respectively form the corresponding super-segment, and these options for selecting the encoding frame duration for a particular segment are limited by the boundaries of the super-segment to which the specified segment belongs . 10. The method according to claim 7, in which each segment of the specified audio signal has a duration of 20 ms and four consecutive segments, respectively, form the corresponding super-segment, while the specified encoding model with conversion allows you to use the duration of the encoding frame 20, 40 and 80 ms, and these options the choice of the length of the coding frame for the segment is limited by the boundaries of the super-segment to which the specified segment belongs. 11. The method according to claim 1, wherein said at least one control parameter comprises an indicator indicating whether to use a shorter or longer coding frame duration, while indicating that a shorter coding frame duration should be applied, eliminates the option of at least as long as the encoding frame is the longest, and an indication that a longer encoding frame is to be used excludes the option of at least the smallest encoding frame. 12. A module for supporting coding of an audio signal, in which at least one segment of the audio signal must be encoded using an encoding model that allows the use of different durations of an encoding frame, wherein said module comprises;
a parameter selection unit configured to determine at least one control parameter at least in part based on the characteristics of said audio signal; and
a frame duration selection unit configured to limit the selection of possible encoding frame durations for at least one segment by means of at least one control parameter provided by said first estimation unit. 13. The module according to item 12, which determines the specified at least one control parameter based on at least one of the following parameters:
an indicator of a short frame, which is determined at least based on the spectral distance; and
the number of selected linear predictor frames with an algebraic code excitation contained in a superframe. 14. The module of claim 12, wherein said block for selecting a frame duration is configured to encode the specified at least one segment using each of the remaining remaining encoding frame lengths with conversion if more than one choice of possible frame durations remains after this limitation encoding, as well as decode the indicated encoded segments again using the corresponding used encoding frame with conversion and select for the specified shey least one segment coding frame length which results in the best decoded audio signal in said at least one segment. 15. The module of claim 14, wherein said block for selecting a frame duration is capable of determining an encoding frame duration that results in a better decoded length by comparing the signal-to-noise ratio obtained for each of said encoding frame durations. 16. The module according to clause 15, in which to determine the specified signal-to-noise ratio for an audio signal obtained at a particular coding frame duration, said frame duration selection unit is configured to first determine a segmented signal-to-noise ratio separately for a plurality of subframes in a corresponding frame encoding and averaging said segmented signal-to-noise ratios of said subframes of a coding frame over an entire coding frame to obtain said signal-to-noise ratio for said at least one segment. 17. The module of claim 12, wherein said parameter selection unit is configured to determine, at least for some segments of the audio signal, based on the characteristics of the audio signal for the corresponding segment of the audio signal, whether to use the specified encoding model or another encoding model, and the ability to provide an indication of the segments for which the specified other coding model was selected as one of the specified control parameters. 18. The module of claim 17, wherein said encoding model is a transform encoding model, and said other encoding model is a linear prediction algorithm with an algebraic code excitation algorithm. 19. An electronic device that contains a support module for encoding an audio signal, in which at least one segment of the specified audio signal must be encoded using an encoding model that allows you to use different durations of the encoding frame, while this module contains:
a parameter selection unit configured to determine at least one control parameter at least in part based on the characteristics of said audio signal; and
a frame duration selection unit, which is configured to limit the selection of possible encoding frame durations for at least one segment by means of at least one control parameter provided by said first estimation unit. 20. The electronic device according to claim 19, which determines the specified at least one control parameter based on at least one of the following parameters:
an indicator of a short frame, which is determined at least based on the spectral distance; and
the number of selected linear predictor frames with an algebraic code excitation contained in a superframe. 21. The electronic device according to claim 19, wherein said block for selecting a frame duration is configured to encode the specified at least one segment using each of the remaining remaining encoding frame lengths with conversion, if after this restriction there is more than one option for choosing possible frame durations encoding, and again decode the specified encoded segments using the corresponding used encoding frame with conversion and choose for bound to at least one segment coding frame length which results in the best decoded audio signal in said at least one segment. 22. The electronic device according to item 21, wherein said block for selecting a frame duration is configured to determine the length of the encoding frame, which leads to a better decoded segment, by comparing the signal-to-noise ratio obtained for each of the indicated encoding frame durations. 23. The electronic device according to item 22, in which to determine the specified signal-to-noise ratio for an audio signal obtained at a particular coding frame duration, said frame duration selection unit is configured to first determine a segmented signal-to-noise ratio separately for a plurality of subframes in the corresponding encoding frame and average the indicated segment signal-to-noise ratios of the indicated subframes of the encoding frame for the whole encoding frame to obtain the specified signal-to-noise ratio for the specified at least one segment. 24. The electronic device of claim 21, wherein said parameter selection unit is configured to determine, at least for some segments of the audio signal, based on the characteristics of the audio signal for the corresponding segment of the audio signal, whether or not to use said encoding model or other encoding model, and with the ability to provide as one of the specified control parameters an indication of the segments for which the specified other coding model was selected. 25. The electronic device according to paragraph 24, wherein said encoding model is a transform encoding model, and said other encoding model is a linear prediction algorithm with an algebraic code. 26. The electronic device according to paragraph 24, in which each segment of the specified audio signal has a predetermined duration and the specified parameter selection unit is capable of providing an indication of the segments for which the specified other coding model has been selected in the corresponding super segment containing a predetermined number of these segments. 27. The electronic device according to claim 19, in which each segment of the specified audio signal has a predetermined duration, and a predetermined number of consecutive segments respectively forms a corresponding super-segment, and said block for selecting a frame duration is capable of limiting the options for selecting the encoding frame duration for a particular segment based on The boundaries of the super-segment to which the specified segment belongs. 28. The electronic device according A.25, in which each segment of the specified audio signal has a duration of 20 ms, and four consecutive segments, respectively, form a super-segment, while this encoding model with conversion allows you to use the duration of the encoding frame 20, 40 and 80 ms and the specified block the choice of the duration of the frame is able to limit the options for choosing the duration of the coding frame for the segment based on the boundaries of the super-segment to which the specified segment belongs. 29. The electronic device according to claim 19, wherein said parameter selection unit is configured to provide as one of the specified control parameters an indicator indicating whether to use a shorter or longer coding frame duration, while indicating that it is necessary to apply a shorter the length of the encoding frame excludes the option of at least the longest duration of the encoding frame, and the indication that you need to apply a longer duration of the encoding frame excludes the option of at least my shorter encoding frame duration. 30. A system for encoding sound, which contains the module according to item 12 and a decoder for decoding audio signals encoded with a variable length of the encoding frame. 31. The system of claim 30, which comprises determining at least one control parameter at least partially based on the characteristics of said audio signal. 32. The system of clause 30, which includes limiting these options for choosing the possible duration of the encoding frame through the specified at least one control parameter. 33. The system according to p. 31, which further comprises, if after the specified restriction there is more than one option to select the possible duration of the encoding frame, encoding the specified at least one segment using each of the indicated remaining durations of the encoding frame with conversion;
decoding said encoded segments with the corresponding used encoding frame length with conversion and
selection for the specified at least one segment of such a length of the encoding frame, which leads to the best decoded audio signal on the specified at least one segment. 34. A software product that stores the code for the audio encoding support program, in which at least one segment of the audio signal must be encoded using an encoding model that allows the use of various encoding frame durations, while the specified program code, when executed in the processing part of the encoder, following operations:
determining at least one control parameter at least partially based on the characteristics of said audio signal; and
the limitation of these options for choosing the possible duration of the coding frame for the specified at least one segment by means of the specified at least one control parameter.

Images