WO2013159364A1

WO2013159364A1 - Method for repairing and decoding air interface voice frame, and signal source side information acquisition method and device

Info

Publication number: WO2013159364A1
Application number: PCT/CN2012/074930
Authority: WO
Inventors: 唐欣; 孙素梅; 乌玟悌谢塔卡谢; 陈炳辉; 吴可镝
Original assignee: 华为技术有限公司; 新加坡科技研究院
Priority date: 2012-04-28
Filing date: 2012-04-28
Publication date: 2013-10-31
Also published as: CN104081669B; CN104081669A

Abstract

Embodiments of the present invention relate to the field of communications, and provide a method for repairing and decoding an air interface voice frame, and a signal source side information acquisition method and device, which can improve the bit error rate and frame error rate performance. The method comprises: acquiring parameters formed of decoding bits; performing reliability calculation on the parameters to obtain soft information corresponding to the parameters and a parameter set containing the parameters which have been corrected using the soft information; performing first cyclic redundancy check on the corrected parameter set; if the first cyclic redundancy check fails, flipping a parameter satisfying a bit flipping rule in the parameter set after determining that the parameter set satisfies a preset rule, so as to obtain a flipped parameter set; performing second cyclic redundancy check on flipped parameter set; and if the second cyclic redundancy check is successful, saving and outputting the flipped parameter set. The embodiment of the present invention is used for joint decoding of signal source channels.

Description

Air interface voice frame repair decoding method, source side information acquisition method and device

The present invention relates to the field of communications, and in particular, to an air interface voice frame repair decoding method, a source side information acquiring method, and a device.

Background technique

With the continuous improvement of multimedia technology, users are increasingly demanding multimedia services such as audio and video. With the diversification of multimedia devices, the quality requirements of users for multimedia are also increasing, which leads to multimedia transmission. The requirements are also increasing. Existing multimedia transmission systems usually use a check code, such as a CRC (Cyclic Redundancy Check) check code, to detect the correctness of the received bit stream.

For example, if the correctness of the received bit stream is detected by using the CRC check code, if the CRC check is incorrect, the BFI (Bad Frame Indication) of the frame is set to 1, and the entire frame is discarded, and then the source is Decoding uses the correlation between adjacent frames to error conceal the wrong frame. Although the error concealment processing can replace or extrapolate the information of the bad frame with the good frame of the previous frame or several frames, the subjective feeling of the degraded speech file after the error concealment is still significantly lower than that of the original speech file. , can not meet the user's requirements for multimedia quality, especially the high fidelity demand for voice.

In order to avoid the adverse effects caused by such full frame discarding, the prior art has proposed a BPI (Bad Parameter Indication) mechanism, although the test results show that the scheme can be effectively improved compared with the BFI mechanism described above. The subjective quality of speech, but it is not compatible with current protocol architectures.

Therefore, in order to solve the above problem, the prior art has proposed some solutions for repairing air interface voice frames, but its improvement on voice quality is limited.

In summary, the air interface speech frame repair and decoding method in the prior art has limited performance improvement on the bit error rate and the frame error rate, and the subjective quality of the speech is not ideal, and there is room for improvement.

Summary of the invention

Embodiments of the present invention provide an air interface voice frame repair decoding method, a source side information acquiring method, and a device, which can improve the bit error rate and the frame error rate performance, thereby effectively improving the subjective quality of the voice. To achieve the above objective, the embodiment of the present invention adopts the following technical solutions: On the one hand, an air interface voice frame repair and decoding method is provided, including:

Obtaining a parameter consisting of decoding bits; performing parameter reliability calculation on the parameter, obtaining soft information corresponding to the parameter, and setting a parameter set by using the soft information to correct the parameter;

Performing a first cyclic redundancy check on the modified parameter set; if the first cyclic redundancy check fails, determining that the parameter set meets a preset rule, and matching bits in the parameter set Flipping the parameters of the flip rule to flip, and obtaining the set of parameters after the flipping;

Performing a second cyclic redundancy check on the inverted parameter set; if the second cyclic redundancy check is successful, saving and outputting the inverted parameter set.

In one aspect, a method for acquiring source side information is provided, including:

Obtain an initial parameter transition probability according to the mean initialization or training corpus;

Calculating a target parameter transition probability according to the initial parameter transition probability, and updating the target parameter transition probability calculated by the current frame to the parameter transition probability of the next frame in real time, so that the air interface voice frame repair decoding device is configured according to each The parameter transition probability obtained in real time by one frame performs parameter reliability calculation on the parameter.

In another aspect, an air interface voice frame repair decoding apparatus is provided, including:

An obtaining module, configured to acquire a parameter composed of decoding bits;

a decoding module, configured to perform parameter reliability calculation on the parameter acquired by the obtaining module, obtain soft information corresponding to the parameter, and obtain a parameter set that is corrected by using the soft information;

a verification module, configured to perform a first cyclic redundancy check on the modified parameter set obtained by the decoding module;

a bit flipping module, configured to: if the first cyclic redundancy check fails of the check module, after determining that the parameter set meets a preset rule, perform a bit that meets a bit flip rule in the parameter set Flip, get the set of parameters after flipping;

a verification module, configured to perform a second cyclic redundancy check on the inverted parameter set obtained by the bit flip module; If the second cyclic redundancy check of the verification module is successful, the virtual source decoder saves, and the first output module outputs the inverted parameter set.

In another aspect, a source side information acquiring device is provided, including:

An obtaining module, configured to obtain an initial parameter transition probability according to the mean initialization or the training corpus;

a calculation module, configured to calculate a target parameter transition probability according to the initial parameter transition probability, and update the target parameter transition probability calculated by the current frame to the parameter transition probability of the next frame in real time, so that the air interface voice frame is repaired The decoding device performs the parameter reliability of the parameter according to the parameter transition probability obtained in real time in each frame. The air interface voice frame repair and decoding method, the source side information acquiring method and device, and the air interface voice frame provided by the embodiments of the present invention are provided. Repairing the decoding device to obtain parameters consisting of decoding bits; performing parameter reliability calculation on the parameters to obtain soft information corresponding to the parameters and the modified parameter set; performing cyclic redundancy check on the modified parameter set; After determining that the parameter set meets the preset rule, flipping the parameter that meets the bit flip rule in the parameter set, and obtaining the inverted parameter set; performing cyclic redundancy check on the inverted parameter set; if successful, saving And output the inverted parameter set. In this way, the parameter with higher error probability, that is, the parameter with lower absolute value of the soft information can be flipped, the pass rate of the CRC check is improved, the bit error rate and the frame error rate performance are improved, and the subjective quality of the voice is improved. .

DRAWINGS

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the embodiments or the description of the prior art will be briefly described below. Obviously, the drawings in the following description are only It is a certain embodiment of the present invention, and other drawings can be obtained from those skilled in the art without any creative work.

1 is a schematic flow chart of a method for repairing and decoding an air interface voice frame according to an embodiment of the present invention;

FIG. 2 is a flowchart of another air interface voice frame repair and decoding method according to an embodiment of the present invention; Schematic diagram

3 is a schematic flowchart of still another method for repairing and decoding an air interface voice frame according to an embodiment of the present invention;

4 is a schematic diagram of BER performance simulation of SBSD and ISCD algorithms after BF flipping according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of FER performance simulation after BF flipping of SBSD and ISCD algorithms according to an embodiment of the present invention; FIG.

FIG. 6 is a schematic diagram of MOS performance simulation of ISCD performing BF inversion and not performing BF flipping according to an embodiment of the present invention; FIG.

FIG. 7 is a schematic diagram of MOS performance simulation of SBSD performing BF flipping and BF flipping without BFSD according to an embodiment of the present invention; FIG.

FIG. 8 is a schematic diagram of BER performance simulation of ISCD performing BF flipping under different SSI estimators according to an embodiment of the present invention; FIG.

FIG. 9 is a schematic diagram of MOS performance simulation after ISCD performing BF flipping under different SSI estimators according to an embodiment of the present invention; FIG.

FIG. 10 is a schematic diagram of BER performance simulation after BF flipping of SBSD under different S SI estimators according to an embodiment of the present invention; FIG.

FIG. 1 is a schematic diagram of MOS performance simulation after BF flipping of SBSD under different S SI estimators according to an embodiment of the present invention;

FIG. 12 is a schematic diagram of MOS performance simulation under the IEC and SEC mechanisms according to an embodiment of the present invention; FIG.

FIG. 13 is a schematic diagram of UER performance simulation after BF flipping of SBSD and ISCD algorithms according to an embodiment of the present invention; FIG.

FIG. 14 is a schematic flowchart of a source side information acquisition method according to an embodiment of the present invention; FIG. 15 is a schematic structural diagram of an air interface voice frame repair and decoding apparatus according to an embodiment of the present invention;

FIG. 16 is a schematic structural diagram of another air interface voice frame repair and decoding apparatus according to an embodiment of the present invention;

FIG. 17 is still another air interface voice frame repair and decoding device according to an embodiment of the present invention. Schematic;

FIG. 18 is a schematic structural diagram of another air interface voice frame repair and decoding device according to an embodiment of the present invention;

FIG. 19 is a schematic structural diagram of still another air interface voice frame repair and decoding apparatus according to an embodiment of the present invention;

FIG. 20 is a schematic structural diagram of a source side information acquiring device according to an embodiment of the present invention.

detailed description

The technical solutions in the embodiments of the present invention are clearly and completely described in the following with reference to the accompanying drawings in the embodiments of the present invention. It is obvious that the described embodiments are only a part of the embodiments of the present invention, but not all embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present invention without creative efforts are within the scope of the present invention.

The air interface voice frame repair and decoding method provided by the embodiment of the present invention is as shown in FIG. 1 , and the method steps include:

S101. The air interface voice frame repairing device acquires a parameter composed of decoding bits.

Exemplarily, assume that the parameter V consists of N decoded bits, so the kth frame parameter = the corresponding binary bit

, where ev,v = {0 ₍ 1, 3'^ - 1}

S102. The air interface speech frame repairing device performs parameter reliability calculation on the parameter, obtains soft information corresponding to the parameter, and sets the parameter after the parameter is corrected by using the soft information.

Further, the air interface voice frame repair device can use ISCD (Iserative Source-Channel Decoding, Joint Source Channel Decoding) and SBSD (Soft-Bit) in JSCD (Joint Source-Channel Decoding). Source Decoding soft bit source decoding) Two algorithms perform parameter reliability calculation to obtain the soft information corresponding to the parameter, and use soft information to modify the parameter or some of the parameters, such as AMR (Adaptive Multi-Rate, Adaptive multi-rate) The A substream in the encoding is corrected.

Further, the air interface voice frame repair device can also use the JSCD algorithm according to each The parameter transition probability obtained by real-time acquisition of one frame is obtained by the posterior probability corresponding to the parameter.

For the ISCD algorithm, in the first iteration, the initial prior information of the channel decoding can be calculated according to the parameter transition probability P: C3⁄43⁄4- and the posterior probability of the previous frame and sent to the interleaver; Perform channel decoding and deinterleave to obtain channel soft information and a priori information of source decoding

The source decoder can calculate the posterior probability ^ρ ί and the soft information ^ί ® according to the above parameters, wherein the posterior probability of the parameter represents the parameter reliability.

The air interface speech frame repair device performs CRC check on the frame parameter. If the check fails, the next iteration calculation is performed. As with the above steps, it is necessary to change that the a priori information of the channel decoding becomes ^£ ©) = ^ ^Μ - (4 (03, if the number of iterations exceeds the preset maximum value, such as 10 times, iterative calculation is no longer performed; if the CRC check is successful, the output makes the posterior probability ^ 1 according to the principle of maximum posterior probability The largest parameter

^{3⁄4W j} ^ refers to the set of bits corresponding to the parameters involved in JSCD correction, such as

The set of A substream bits, the set of bits corresponding to the remaining parameters, such as the B substream bit set and the C substream bit set, are outputted by channel decoding.

For the SBSD algorithm, the CRC check can be performed on the parameters of the current subframe before the algorithm is enabled. Since the SBSD algorithm is similar to an error concealment technique, the algorithm can be started when the CRC check fails. The source channel joint decoding device calculates the error probability of the ith bit parameter of the kth frame according to the bit soft information output by the channel decoding ((0) according to the error probability, obtains the bit transmission probability The corpse [4 (o/( (o), then calculate the sequence transmission probability of the transmission sequence ^k to the decoding sequence ^ = ( .(3⁄4) (1)'"': - 1)), where

P{h

Then according to Ρ( ), use SSI ( Side Source

Side Information, source side information) The transfer parameter transfer probability ^P3⁄4 - ^i;i and

And using parameters

C

The posterior probability represents the parameter reliability. among them

Winter search

For the posterior probability test of the SBSD algorithm parameters, according to the principle of maximum posterior probability, the output makes the posterior probability rate

^Ρ The bit sequence () of the binary representation corresponding to the largest parameter, ie IS _s S _s I _r , is the set of bits corresponding to the parameters involved in the JSCD correction, such as the A substream bit set, and the remaining parameters. The set of bits, such as the B substream bit set and the C substream bit set, are channel decoded output.

S103. The air interface voice frame repairing device performs a first C R C check on the modified parameter set.

Exemplarily, the air interface speech frame repair device may perform a first CC check on the entire modified parameter set, or may perform a first CRC check on the modified parameter in the modified set, as modified in the modified AMR code. The A substream performs a first CRC check.

It should be noted that this embodiment lists GSM (Global System for AMR business under Mobile communications, Global System for Mobile Communications. The bits that make up the parameters in the AMR encoding are divided into A\B\C substreams according to their importance, where the A substream is of the highest importance and is protected by the CRC check. In the SEC (Standard Error Concealment) mechanism defined by 3GPP (The 3rd Generation Partnership Project), if the scorpion stream has an error, the frame is considered to be a bad frame. All are discarded. Therefore, in this embodiment, the CRC detection is continued after the A substream is modified, so that when the verification fails, the probability of bit errors in the A substream is reduced by inverting the decoded bits. However, the A bit stream of the AMR in the GSM system is used as an example of the modified bit set as an example. However, without limitation, the modified bit set in other systems or other services is within the protection range.

S104. If the first CRC check of the air interface voice frame repair device fails, after determining that the modified parameter set meets the preset rule, the bit that meets the bit flip rule in the parameter set is flipped, and the parameter set after the rollover is obtained. . It should be noted that, assuming that the soft information of the bit in the parameter set is ^£ ^ ⁽ , the error probability of the bit is: ― ^1÷ ^* ^{ί3⁄4ί ΐ:}

The larger the absolute value, the lower the error probability of this bit. The error parameters can be further corrected by flipping those bits with a higher probability of error (lower absolute soft information) until the CRC check passes.

For example, if the air interface speech frame repair device determines that the modified parameter set meets the preset rule, for example, determining the absolute value of the soft information corresponding to the bit in the parameter in the A substream or the bit corresponding to the other parameters in the modified parameter set. If the absolute value of the soft information is less than or greater than a preset threshold, the BF (Bit-Flipping) algorithm is started to perform BF flipping of the bits in the A substream that conform to the bit flip rule, such as the A substream. The bit whose absolute value of the soft information corresponding to the middle bit is smaller than the preset threshold value is subjected to BF inversion.

S105. The air interface voice frame repairing device performs a second CRC check on the inverted parameter set. A second CRC check is performed on the flipped parameter set, and the flipped parameter set includes parameters consisting of inverted bits and parameters composed of unflipped bits.

It should be noted that, in the parameter set of the flipping, the bit satisfying the above-mentioned bit flipping rule has been flipped, and the bit that does not satisfy the rule is not flipped, and the air interface voice frame repairing device includes some flipping for the inverted parameter set. The A substream of the bit is subjected to a second CRC check.

S 106. The air interface voice frame repairing device saves and outputs the inverted parameter set if the second CRC check succeeds.

Further, the air interface voice frame repairing device performs the second CRC check successfully, and can also save and output the inverted parameter set and the posterior probability corresponding to the parameter in the parameter set.

The method for jointly decoding a source channel according to an embodiment of the present invention, the air interface voice frame repair device acquires a parameter composed of decoding bits; performs parameter reliability calculation on the parameter to obtain soft information corresponding to the parameter and the modified parameter set; The modified parameter set performs cyclic redundancy check; if it fails, after determining that the parameter set meets the preset rule, the bit that meets the bit flip rule in the parameter set is flipped to obtain the inverted parameter set; The parameter set performs cyclic redundancy check; if successful, saves and outputs the inverted parameter set. In this way, by translating the bit with a higher error probability, that is, the lower absolute value of the soft information, the pass rate of the CRC check is improved, the bit error rate and the frame error rate performance are improved, and the subjective quality of the voice is improved. .

The air interface voice frame repair and decoding method provided by the embodiment of the present invention is based on the decoding of the AMR service in the GSM, wherein the parameter set modified by the soft information includes a parameter composed of the first type of bits and a parameter composed of the second type of bits. The first type of bit is a CRC bit, and the second type of bit is a bit in the A substream, which is only illustrated by way of example, but is not limited thereto. The method steps are as shown in FIG. 2, including:

S201. The air interface voice frame repair device acquires a parameter composed of decoding bits.

S202. The air interface voice frame repairing device acquires the posterior probability and the soft information corresponding to the parameter through the ISCD and the SBSD according to the parameter, and obtains the parameter set corrected by the soft information.

The modified parameter set is composed of a parameter consisting of CRC bits and an A substream. It should be noted that if the SBSD algorithm is adopted, the CRC can be performed after the parameters are acquired. Verification, if the verification fails, the SB SD algorithm is started.

Further, if the SBSD algorithm is used, the bit information used after step S204 is inverted and the parameters are

Posterior probability is

Further, since the posterior probability corresponding to the parameter can be calculated according to the parameter transition probability obtained in real time, if the voice signal to be transmitted is known, the parameter transition probability obtained by the SSI (Source Side Information) is Ρ^ ^ - ^ can be obtained by calculation of the number of transmissions. Since the speech signal is generally a non-stationary signal, the redundancy of adjacent frames is much larger than the inter-frame redundancy that is far apart. Therefore, using the parameters of the previous frame and the frame parameters to calculate the number of transmissions can reduce the impact of redundancy. , assuming the previous frame decoding parameter - i = the current frame decoding parameter ^{= γ} , if the frame CRC check does not pass, then the frame parameter: = ν has a posterior probability that is not ι. Therefore, the number of transmissions ^& 03⁄4/3⁄4-i] needs to be added =

EF. In order to reduce the storage complexity, the following algorithm can be used:

Using a mean initialization or training corpus to obtain a set of initialization parameter transition probabilities 3⁄4 3⁄4-i), wherein the training corpus is obtained by statistically analyzing speech samples of people of different races, genders, ages, etc., and passing the speech sample file. The sample file is statistically calculated, and the initialization parameter transition probability is obtained by performing a desired calculation or the like.

The target parameter transition probability is obtained according to the initial parameter transition probability. The target parameter transition probability is the parameter transition probability obtained in real time after the end of each frame decoding, that is, after each frame decoding is completed, the obtained parameter transition probability is updated to the previous parameter transition probability. ^p :

Two

Where ⁿ weighting factor, 3⁄4r' (3⁄4= W << = 》, can normalize the parameter transition probability.

It is worth noting that the S SI scheme proposed by the prior art is based on the known transmission voice signal, which is also called off-line S SI. Therefore, the scheme refers to the above-mentioned SSI scheme which is dynamically updated in real time as on-line. SSI„

S203. The air interface voice frame repair device performs a first CRC check on the modified parameter set. It should be noted that if the first CRC check succeeds, step S204 is performed, and if the first CRC check fails, step S205 is performed.

5204. The air interface speech frame repair device selects a parameter corresponding to the maximum posterior probability, and outputs the selected parameter.

It should be noted that, since the first CRC check of the current frame parameter is successful, the output current frame information may be stored in the VSD (Virtual Source Decoder) as good frame information, so that the air interface voice frame is obtained. The repair device can use the stored parameters to perform error concealment processing on the error parameters of the bad frame when the next frame is a bad frame.

5205. The air interface voice frame repairing device sorts the absolute values of the soft information corresponding to the bits of each parameter in the A substream in the modified parameter set by small to large.

Incidentally, the absolute value of the soft information may be A sub streams corresponding to each bit _{£ ₍} '^ _₍₀₎ | sorted in descending order or not sorted, where only small to large For the preferred embodiment, it is not limited thereto.

If the preset rule is met, S206 is executed; otherwise, S209 is executed.

Exemplarily, it can be determined that the modified parameter set conforms to the preset rule when the following three conditions are met:

If the absolute value of the soft information corresponding to the CRC bit is greater than the second preset threshold L _th , that is: min "¹; and the soft information corresponding to each bit in the A substream sorted from small to large is absolutely The absolute value of the first M soft information in the value is also greater than the second preset threshold, where M=L _mp +l , and M is an integer smaller than the total number of bits included in the A substream, namely: pre-L _mp

+1 ^ ¹ ' At the same time, the absolute value of the smallest soft information corresponding to the bit in the A substream is greater than the third preset threshold, namely: min " ¹ .

5206. The ratio of the absolute value of the soft information to the absolute value of the minimum soft information in the parameter set of the air interface voice frame repairing device is smaller than the bit in the A substream of the fourth preset threshold, and the third parameter set containing the inverted bit is performed. CRC check.

It is worth pointing out that the absolute value of the soft information corresponding to the bits in all the dice streams The ratio of the absolute value to the minimum soft information, _Aa is the fourth preset threshold, and the air interface speech frame repair device flips (3⁄4 < 3⁄4 bits).

If the third CRC check performed after the flip is successful, S207 is performed, and if the third CRC check fails, S210 is performed.

S 207. The air interface voice frame repairing device determines the validity of the parameter set according to the effective range.

Exemplarily, the effective range can be set according to whether the parameter is valid. For example, the adaptive codebook gain parameter in the AMR code has 6 bits, that is, a total of ^{23⁄4 = δ4: the} value:

0-63, but the actual encoding can only be up to 60, so the maximum range is 0~60, so the effective range is 0~60. If the adaptive codebook gain parameter is exceeded in the parameter set after flipping

60, it is considered untrustworthy, that is, it is judged that this parameter set is invalid. If the adaptive codebook gain parameter in the parameter set does not exceed 60, it is judged that this parameter set is valid. It is worth noting that the validity of the parameter range is determined by the effective range of the adaptive codebook gain parameter in the AMR coding. The validity range of other parameters or the effective range of the current system configuration can also be used to judge the validity. Do this with any restrictions.

If it is judged to be valid, S208 is executed, otherwise, S209 is executed.

5208. The air interface voice frame repair device outputs a parameter in the parameter set of the flip and an updated posterior probability corresponding to the parameter.

Further, since the CRC check of the current frame parameter is successful, and it is determined that the parameter set is valid, the output current frame information may be stored in the VSD (Virtual Source Decoder) as good frame information, The air interface voice frame repairing device can perform error concealment processing for the error parameter of the next frame according to the current frame information.

5209. The air interface voice frame repair device outputs the JSCD modified parameter set and the posterior probability corresponding to the parameter.

Among them, the modified parameter of JSCD is the posterior probability corresponding to the parameter is ^P (D.

5210. The parameters of the A substream soft information whose absolute value is smaller than the fifth preset threshold in the parameter set of the air interface voice frame repairing device flipping correction, and the parameter set including the flipped parameter is performed. The second CRC check' is the parameter of the flip correction, and (d))i

A substream soft information absolute value, r _th is the fifth preset threshold.

If the CRC check succeeds after the rollover, S207 is executed, and if the check fails, S209 is executed. It is worth noting that the selection of each of the above thresholds needs to consider the trade-off between FER (Frame Error Rate) and UER (Undetected Error Rate). UER indicates the probability that the CRC check is successful and the A substream still has a bit error. Table 1 shows a set of threshold values. Each of the above thresholds can be selected and set from the range of values.

Table 1

It should be noted that, as shown in FIG. 3, after the step S205, the inversion of S206 and S210 is not limited, and S210 may be performed first, that is, if the preset rule is met, then S210 is performed, otherwise, S209 is performed; Step S210; After the inversion, if the fourth CRC check is successful, then S207 is performed. If the fourth CRC check fails, step S206 is performed. After the step S206 is reversed, if the second CRC check is successful, step S207 is performed. Otherwise, Step S209 is performed.

It is worth noting that the first CRC check, the second CRC check, the third CRC check, and the fourth CRC check only indicate that the set of parameters for the check is different, or the conditions for verifying start are different, but All are CRC check, no restrictions are imposed.

As shown in FIG. 4 and FIG. 5, after the BFSD and ISCD algorithms are used in the air interface speech frame repairing device, the BER and FER of the A substream are significantly reduced, and a gain of about 0.5 dB can be obtained. The performance of ISCD-BF is better than that of SBSD-BF in dry ratio. With the increase of carrier-to-interference ratio, the performance of the two algorithms is relatively close and superior.

In addition, Figures 6 and 7 show that JSCD with BF algorithm can obtain better MOS. (Mean Opinion Score, average subjective score) Performance, Similarly, the MOS score performance of ISCD-BF is better than SBSD-BF at low carrier to dry ratio.

Figure 8 and Figure 9 show the BER performance and MOS performance of the A substream using different SSI estimators under the ISCD-BF algorithm. By comparing the performance of the BER in Figure 8 with the performance of the MOS in Figure 9, the performance of the on-line SSI estimator is close to the ideal off-line SSI estimator performance. However, there may be a performance backplane at high carrier to dry ratios. Figure 10 and Figure 1 show the BER performance and MOS performance of the A substream using different SSI estimators under the SBSD-BF algorithm. As can be seen from the figure, the performance of each SSI estimator is basically close under SBSD-BF, and SBSD-BF does not exhibit the performance backplane appearing in ISCD-BF at high carrier-to-interference ratio. This is because the SBSD-BF algorithm only starts when the CRC check error occurs, and the probability of a CRC check error at the high load-to-interference ratio is very low.

Further, after step S209, the air interface voice frame repairing device performs step S211 to perform IEC (Individual Error Concealment) on the error parameter.

The air interface voice frame repairing device identifies the parameter whose posterior probability is less than the first preset threshold P _th as a bad parameter.

Exemplarily, the air interface speech frame repair device may set the number of bad parameters of the current frame, for example, set to 4. If the number of bad parameters of the frame is less than 4, select the minimum 4 parameters of the posterior probability ' ^{Ρ: ί} ^ ) The parameter is a bad parameter. For example, when there are only 3 bad parameters in the frame, the parameter with the smallest thickness parameter among other good parameters is also selected as a bad parameter, which can reduce the probability of noise generation.

It should be noted that the first preset threshold may be any value selected between 0-2.

5212. The air interface voice frame repairing device updates the bad parameters by using parameters in the pre-stored good frame. It should be noted that the parameters in the pre-stored good frame can be pre-stored in the VSD, wherein the pre-stored good frame is a frame that does not identify a bad parameter, and the air interface voice frame repair device can perform the current frame according to the parameters of the previous good frame. If the bad parameters are updated, the parameter similarity rate of the adjacent frames may be higher, which is convenient for improving the voice quality of the current frame after updating the bad parameters, but not all bad previous frames are good frames, and only the case is illustrated here. , do not make any restrictions on this.

For example, the air interface speech frame repair device may use the LSF (Linear Spectrum Frequency), the adaptive codebook gain parameter, and the fixed codebook gain parameter in the previous good frame to update the bad parameters in the current bad frame. . For example, the LSF and LSF mean values of the current frame in which the bad parameters are replaced are used by using the LSF of the previous good frame and the LSF mean; and the pre-stored good frames are reused. The signal amplitude limits the adaptive codebook gain parameter and the fixed codebook gain parameter of the current frame, such as limiting the adaptive codebook gain parameter to 10, and limiting the fixed codebook gain parameter to 20; or The adaptive codebook gain parameter is limited to 13 and the fixed codebook gain parameter is limited to 22, which is only illustrated by the above numerical values, and is not limited thereto.

Preferably, the air interface speech frame repairing device can further replace the pitch delay parameter at the corresponding odd position of the current frame where the bad parameter is located by using the pre-stored pitch delay parameter of the odd position in the previous good frame, and use the pre-stored previous good frame. The pitch delay parameter of the even position in the middle adds the preset offset and replaces the pitch delay parameter at the corresponding even position of the current frame. It may be assumed that the current frame contains four subframes, and since the second subframe and the fourth subframe are less redundant, the second subframe and the fourth subframe of the previous good frame may be used instead, and the first subframe is replaced by the first subframe. And the third subframe can use the gene delay parameter of the first subframe of the previous good frame plus a preset offset, such as the gene delay parameter of the first subframe is 10, the preset offset If the quantity is 2, the gene delay parameter of the first subframe of the current frame is 12, and the gene delay parameter of the fourth subframe can be obtained by the same reason.

Figure 6, Figure 7, and Figure 12 show that the IEC mechanism implemented in steps S211 and S212 can further improve the performance of the algorithm compared to the SEC (Standard Error Concealment) mechanism in the prior art.

Due to the BF flipping, the possibility of CRC miss detection is increased, and the missed detection increases the risk of introducing impurities for the voice service. Therefore, it is necessary to further compare the UER (Undetected Error Rate) of the present embodiment and the existing scheme. ). Figure 13 shows that the UER of this embodiment is basically the same as the existing scheme, and the UER is not increased by improving the FER and MOS performance.

The method for jointly decoding a source channel according to an embodiment of the present invention, the air interface voice frame repair device acquires a parameter composed of decoding bits; performs parameter reliability calculation on the parameter to obtain soft information corresponding to the parameter and the modified parameter set; The modified parameter set performs cyclic redundancy check; if it fails, after determining that the parameter set meets the preset rule, the bit that meets the bit flip rule in the parameter set is flipped to obtain the inverted parameter set; The parameter set performs cyclic redundancy check; if successful, saves and outputs the inverted parameter set. In this way, the parameter with higher error probability, that is, the parameter with lower absolute value of the soft information can be flipped, the pass rate of the CRC check is improved, the bit error rate and the frame error rate performance are improved, and the subjective quality of the voice is improved. . The source side information obtaining method provided by the embodiment of the present invention is as shown in FIG. 14 , and the method steps include:

S 301. The source side information acquiring device obtains an initial parameter transition probability according to the mean initialization or the training corpus.

Exemplarily, if the speech signal to be transmitted is known, the parameter transition probability ^p c3⁄4 / - obtained by the SSI can be obtained by calculation of the number of transmissions. Since the speech signal is generally a non-stationary signal, the redundancy of adjacent frames is much larger than the inter-frame redundancy that is far apart. Therefore, using the parameters of the previous frame and the frame parameters to calculate the number of transmissions can reduce the impact of redundancy. , assuming the previous frame decoding parameter -i ^{= w} , this frame decoding parameter ^{= v} , if this frame

If the CRC check fails, the posterior probability of this frame parameter 3⁄4 - ^v is not 1. Therefore, the number of transmissions 1) needs to be added =

£ F In order to reduce the storage complexity of ₃ · 3, the source side information acquisition device uses the mean initialization or training corpus to obtain a set of initialization parameter transfer probability ^^^-

S302. The source side information acquiring device calculates the target parameter transition probability according to the initial parameter transition probability, and updates the target parameter transition probability calculated by the current frame to the parameter transition probability of the next frame in real time, so that the air interface voice frame repair decoding device The parameter reliability is calculated based on the parameter transition probability obtained in real time for each frame.

The target parameter transition probability is obtained according to the initial parameter transition probability. The target parameter transition probability is the parameter transition probability obtained in real time after the end of each frame decoding, that is, after each frame decoding is completed, the obtained parameter transition probability is updated to the previous parameter transition probability. ^P:i - ' ^A U:

Where ^η is the weighting factor, ^― — can be normalized to the parameter transition probability.

It should be noted that the calculation of the parameter reliability of the parameters by the air interface voice frame repair and decoding device according to the parameter transition probability obtained in real time in each frame has been expanded in the above embodiments, and details are not described herein again.

The method for acquiring source side information provided by the embodiment of the present invention, the initial parameter transition probability acquired by the source side information acquiring device, and calculating according to the obtained initial parameter transition probability The target parameter transition probability is obtained, and the target parameter transition probability calculated in the current frame is updated in real time to the parameter transition probability of the next frame. In this way, the source side information acquiring device can update the parameter transition probability in real time, and provide a more accurate parameter transition probability for other connected devices, such as an air interface speech frame repair decoding device, which uses the parameter transfer probability with higher correct rate for processing. , in turn to ensure the correct rate of data processing of other connected devices.

The air interface voice frame repairing device 30 provided by the embodiment of the present invention, as shown in FIG. 15 , includes:

The obtaining module 301 is configured to obtain a parameter composed of decoding bits.

Illustratively, the acquisition module 301 can obtain parameters consisting of decoded bits output from the channel decoder.

The calculation module 302 is configured to perform resource source joint decoding calculation parameter reliability calculation on the parameters acquired by the obtaining module 301, obtain soft information corresponding to the parameter, and use the soft information to correct the parameter set.

Further, the parameter set modified by the soft information may be a new parameter set composed of parameters consisting of the A substream and the CRC bit, and is recorded as a modified parameter set, wherein the CRC bit can be used to determine whether the CRC check is trusted. .

The verification module 303 is further configured to perform a first CRC check on the modified parameter set obtained by the calculation module 302.

The bit flipping module 304 is configured to: if the first CRC check fails by the check module 303 on the modified parameter set, and after determining that the modified parameter set meets the preset rule, the bit flip rule is matched in the parameter set. The bits are flipped to obtain a flipped set of parameters containing the inverted bit composition parameters.

It should be noted that the conditions and methods of flipping are described in detail in the above embodiments, and are not described herein again.

The verification module 303 is configured to perform a second cyclic redundancy check on the inverted parameter set obtained by the bit flip module 304.

If the second CRC check by the check module 303 is successful, it is saved by the virtual source decoder 305, and the first output module 306 outputs the inverted set of parameters.

Further, as shown in FIG. 16, the adaptive air interface voice frame repairing device 30 further includes: The determining module 307 is configured to perform validity determination according to the valid range in the actual transmission according to the inverted set of parameters obtained by the bit flip module 304 after the second CRC check by the check module 303 is successful.

It is worth noting that the effective range is limited according to the lowest value and the highest value that can be achieved by the actual coding of the parameter. For example, the adaptive codebook gain in the AMR speech coding occupies 6 bits, and ideally, 64 values should be taken. , recorded as 0~63, but the actual code can only reach 60, so the range is limited to 0~60.

If the determining module 307 determines that the flipped parameter set is within the valid range, the virtual source decoder 305 saves the flipped parameter set, and the first output module 306 outputs the inverted parameter set.

If the judging module 307 determines that the flipped parameter set is outside the valid range, the virtual source decoder 305 saves the corrected parameter set before the flip, and the second output module 308 outputs the corrected parameter set before the flip.

It should be noted that if the second CRC check by the check module 303 fails, the virtual source decoder 305 stores the corrected parameter set before the flip, and the second output module 308 outputs the corrected parameter set before the flip.

Preferably, the virtual source decoder 305 saves, and the first output module 306 outputs the inverted parameter set and the posterior probability corresponding to the parameter in the parameter set.

Or,

The virtual source decoder 305 saves, and the second output module 308 outputs the corrected parameter set before the flip and the posterior probability corresponding to the parameter in the parameter set.

Further, as shown in FIG. 17, the air interface voice frame repairing apparatus 30 further includes: a source side information estimator 309, configured to acquire a parameter transition probability of each frame in real time, so that the calculating module 302 performs parameter reliability on the parameter. Calculate, calculate the posterior probability corresponding to the parameter according to the parameter transition probability.

Preferably, the source side information estimator 309 can also be configured to obtain an initial parameter transition probability according to the mean initialization or training corpus; calculate the target parameter transition probability according to the initial parameter transition probability, and calculate the target parameter transition probability in the current frame in real time. Update to the parameter transition probability of the next frame. Further, as shown in FIG. 18, the air interface voice frame repairing device 30 further includes: an identifying module 310, configured to use the virtual source decoder 305 to save the posterior probability of the modified parameter set before the flipping is smaller than the first preset. The parameter of the threshold is identified as a bad parameter.

The independent error concealment module 31 1 is configured to update the bad parameters by using parameters in the good frame in the pre-stored virtual source decoder 305, wherein the good frame is a frame that does not identify the bad parameter.

Illustratively, the independent error concealment module 31 1 replaces the LSF and LSF mean of the current frame in which the bad parameters are located by using the good frame midline spectral frequency LSF and the LSF mean of the pre-existing virtual source decoder 305; using the pre-existing virtual source decoder 305 The signal amplitude in the good frame limits the adaptive codebook gain parameter and the fixed codebook gain parameter of the current frame.

Exemplarily, the independent error concealment module 31 1 is configured to replace the pitch delay at the corresponding odd position of the current frame where the bad parameter is located, respectively, by using the pitch delay parameter of the odd position in the previous good frame of the pre-existing virtual source decoder 305. The parameter is used to replace the pitch delay parameter of the corresponding even position of the current frame by using the pitch delay parameter of the even position in the previous good frame of the pre-existing virtual source decoder 305 and adding the preset offset.

Further, as shown in FIG. 19, the air interface voice frame repairing apparatus 30 further includes: a maximum a posteriori probability estimator 312, configured to successfully perform a first CRC check on the first CRC check performed by the check module 303, The parameter corresponding to the maximum a posteriori probability in the inverted parameter set is selected, and the selected parameter is output. The virtual source decoder 305 also saves the selected parameter and saves the frame of the selected parameter as a good frame.

The air interface voice frame repairing device 30 can use the air interface voice frame repairing method provided in the foregoing embodiment. The method has been described in detail in the foregoing embodiments, and details are not described herein again.

The air interface voice frame repairing device 30 and the air interface voice frame repairing device 30 of the embodiment of the present invention acquire parameters composed of decoding bits; perform parameter reliability calculation on the parameters to obtain soft information corresponding to the parameters and the modified parameter set; The modified parameter set performs cyclic redundancy check; if it fails, after determining that the parameter set meets the preset rule, the bit that meets the bit flip rule in the parameter set is flipped to obtain the inverted parameter set; The parameter set performs cyclic redundancy check; if successful, saves and outputs the inverted parameter set. In this way, the parameter with higher error probability, that is, the parameter with lower absolute value of the soft information can be flipped, the pass rate of the CRC check is improved, the bit error rate and the frame error rate performance are improved, and the subjective quality of the voice is improved. .

The source side information acquiring device 40 provided by the embodiment of the present invention is as shown in FIG. Includes:

The obtaining module 401 is configured to obtain an initial parameter transition probability according to the mean initialization or the training corpus.

The calculation module 402 is configured to calculate a target parameter transition probability according to the initial parameter transition probability, and update the target parameter transition probability calculated by the current frame to the parameter transition probability of the next frame in real time, so that the air interface voice frame is obtained. The repair decoding device 30 performs the parameter reliability meter on the parameter according to the parameter transition probability obtained in real time for each frame. The source side information acquiring device 40 can use the source side information obtaining method provided by the foregoing embodiment. The method has been described in detail in the above embodiments, and details are not described herein again.

Further, the source side information acquiring device 40 can also be used as the source side information estimator 309 in the above-mentioned air interface voice frame repairing device 30, and details are not described herein.

The source side information obtaining device 40 and the source parameter side acquiring information obtained by the source side information acquiring device 40 according to the embodiment of the present invention calculate the target parameter transition probability according to the obtained initial parameter transition probability, and calculate the current frame in real time. The target parameter transition probability is updated to the parameter transition probability of the next frame. In this way, the source side information acquiring device can update the parameter transition probability in real time, and provide a more accurate parameter transition probability for other devices connected to the source side information acquiring device 40, such as the air interface voice frame repair decoding device 30. A higher rate of parameter transfer probability is processed to ensure the correct rate of data processing of the connected device.

The above is only the specific embodiment of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily think of changes or substitutions within the technical scope of the present invention. It should be covered by the scope of the present invention. Therefore, the scope of the invention should be determined by the scope of the claims.

Claims

Claim

An air interface voice frame repair and decoding method, comprising: acquiring a parameter consisting of decoding bits; performing parameter reliability calculation on the parameter, obtaining soft information corresponding to the parameter, and using the a set of parameters for which the soft information corrects the parameters;

Performing a first cyclic redundancy check on the modified parameter set; if the first cyclic redundancy check fails, determining that the parameter set meets a preset rule, and matching bits in the parameter set The bits of the flip rule are flipped to obtain a set of parameters after the flipping;

The method according to claim 1, wherein, after the second cyclic redundancy check succeeds, the method further includes:

Performing validity determination according to the valid range of the inverted parameter set in the actual transmission; if the inverted parameter set is within the valid range, saving and outputting the inverted parameter set; If the parameter set is outside the valid range, the modified parameter set before the flip is saved and output.

The method according to claim 1, wherein after performing the second cyclic redundancy check on the inverted parameter set, the method further includes:

If the second cyclic redundancy check fails, the modified parameter set before the flip is saved and output.

The method according to any one of claims 1 to 3, wherein after the obtaining the parameter consisting of the decoded bits, the method further includes:

The parameter reliability calculation is performed on the parameter, and the posterior probability corresponding to the parameter is calculated according to the parameter transfer probability obtained in real time for each frame.

5. The method of claim 4, wherein

And saving and outputting the posterior probability corresponding to the parameter in the corrected parameter set before the flipping, when saving and outputting the modified parameter set before the flipping;

or, When the inverted parameter set is saved and output, the posterior probability corresponding to the parameter in the inverted parameter set is also saved and output.

The method according to claim 5, wherein, when saving and outputting the corrected parameter set before flipping, storing and outputting the parameter corresponding to the parameter in the corrected parameter set before flipping After the posterior probability, it also includes:

And identifying, in the set of the modified parameters before the flipping that is saved and output, a parameter whose posterior probability is less than the first preset threshold is a bad parameter;

The bad parameters are updated with parameters in a pre-stored good frame, wherein the good frame is a frame that does not identify a bad parameter.

The method according to claim 1 or 4, wherein the parameter reliability calculation is performed on the parameter, and the posterior probability corresponding to the parameter is calculated according to the parameter transition probability obtained in real time for each frame. , Also includes:

Calculating the target parameter transition probability according to the initial parameter transition probability, and updating the target parameter transition probability calculated by the current frame to the parameter transition probability of the next frame in real time.

8. The method according to claim 6, wherein the updating the bad parameters by using a good parameter in a pre-stored good frame comprises:

And using the pre-stored good frame mid-line spectrum frequency LSF and the LSF mean value to replace the LSF and LSF mean values of the current frame in which the bad parameter is located;

The adaptive codebook gain parameter and the fixed codebook gain parameter of the current frame are limited by the pre-stored signal amplitude in the good frame.

9. The method according to claim 6, wherein the updating the bad parameters by using a good parameter in a pre-stored good frame comprises:

Using the pitch delay parameters of the odd-numbered positions in the pre-stored previous good frame to replace the pitch delay parameters at the corresponding odd positions of the current frame where the bad parameters are located, respectively, and using the pre-stored base delay of the even-numbered positions in the previous good frame. The time parameter is added to the preset offset and then replaced with the pitch delay parameter at the corresponding even position of the current frame.

10. The method of claim 1 wherein: said correcting said After the first parameter set performs the first cyclic redundancy check, it also includes:

If the first cyclic redundancy check succeeds, the parameter corresponding to the maximum posterior probability in the inverted parameter set is selected, and the selected parameter is saved and output.

1 1 . A method for acquiring information of a source side, which is characterized by comprising:

An air interface voice frame repair and decoding device, comprising: an acquiring module, configured to acquire a parameter composed of decoding bits;

a calculation module, configured to perform parameter reliability calculation on the parameter acquired by the acquisition module, obtain soft information corresponding to the parameter, and obtain a parameter set corrected by using the soft information to the parameter;

a verification module, configured to perform a first cyclic redundancy check on the modified parameter set obtained by the source channel joint decoding module;

a bit flipping module, configured to: if the first cyclic redundancy check fails in the check module, after determining that the parameter set meets a preset rule, perform a bit that meets a bit flip rule in the parameter set Flip, get the set of parameters after flipping;

a verification module, configured to perform a second cyclic redundancy check on the inverted parameter set obtained by the bit flip module;

And if the second cyclic redundancy check of the verification module is successful, saved by the virtual source decoder, and the first output module outputs the inverted parameter set.

The device according to claim 12, further comprising: a determining module, configured to: according to the bit flip module, after the second cyclic redundancy check by the check module is successful The set of parameters after the flipping is validated in the effective range of the actual transmission;

If the parameter set after the flipping is within the valid range, the virtual source decoder is saved, and the first output module outputs the inverted parameter set; And if the parameter set after the flipping is outside the valid range, the virtual source decoder is saved, and the second output module outputs the modified parameter set before the flipping.

14. Apparatus according to claim 12 wherein:

And if the second cyclic redundancy check by the check module fails, the virtual source decoder saves, and the second output module outputs the corrected parameter set before flipping.

The device according to any one of claims 12 to 14, further comprising:

a source side information estimator, configured to acquire a parameter transition probability of each frame in real time, so that the calculation module performs parameter reliability calculation on the parameter, and calculates a posteriori corresponding to the parameter according to the parameter transition probability Probability.

16. Apparatus according to claim 15 wherein:

The virtual source decoder saves, the first output module outputs the inverted parameter set and a posterior probability corresponding to the parameter in the parameter set.

Or,

The virtual source decoder saves, the second output module outputs the modified parameter set before flipping and the posterior probability corresponding to the parameter in the parameter set.

The device according to claim 16, further comprising: an identifier module, wherein the posterior probability of the modified parameter set before flipping saved by the virtual source decoder is less than A parameter of a preset threshold is identified as a bad parameter; an independent error concealment module is configured to update the bad parameter by pre-storing parameters in a good frame in the virtual source decoder, where the good frame is unidentified bad The frame of the parameter.

18. Apparatus according to claim 12 or claim 15 wherein:

The source side information estimator is specifically configured to obtain an initial parameter transition probability according to the mean value initialization or training corpus; calculate a target parameter transition probability according to the initial parameter transition probability, and transfer the target parameter calculated in the current frame in real time. The probability is updated to the parameter transition probability of the next frame.

19. Apparatus according to claim 17 wherein:

The independent error concealing module is specifically configured to decode by using the pre-existing virtual source The good frame mid-line spectrum frequency LSF and the LSF mean replace the LSF and LSF mean of the current frame in which the bad parameter is located; the adaptation of the current frame using the signal amplitude in the good frame pre-existing in the virtual source decoder The codebook gain parameter and the fixed codebook gain parameter are limited.

20. Apparatus according to claim 17 wherein:

The independent error concealment module is specifically configured to replace, by using a pitch delay parameter of an odd position in a previous good frame of the virtual source decoder, a pitch delay at a corresponding odd position of the current frame where the bad parameter is located, respectively. a time parameter, using a pitch delay parameter pre-existing in an even position of the previous good frame of the virtual source decoder, adding a preset offset, and respectively replacing the pitch delay parameter at the corresponding even position of the current frame .

The device according to claim 12, further comprising: a maximum a posteriori probability estimator, configured to: if the first cyclic redundancy check performed by the first cyclic redundancy check succeeds, select Outputting the selected parameter to the parameter corresponding to the maximum posterior probability in the inverted parameter set;

The virtual source decoder saves the selected parameters.

22. A source side information acquiring device, comprising:

An obtaining module, configured to obtain an initial parameter transfer probability according to the mean initialization or training corpus;

a calculation module, configured to calculate a target parameter transition probability according to the initial parameter transition probability, and update the target parameter transition probability calculated by the current frame to the parameter transition probability of the next frame in real time, so that the air interface voice frame is repaired The decoding device performs parameter reliability calculation on the parameter according to the parameter transition probability obtained in real time for each frame.