CN101895373A

CN101895373A - Channel decoding method, system and device

Info

Publication number: CN101895373A
Application number: CN 201010234610
Authority: CN
Inventors: 朱佥; 唐欣; 李明; 刘小青
Original assignee: Huawei Technologies Co Ltd
Current assignee: Huawei Technologies Co Ltd
Priority date: 2010-07-21
Filing date: 2010-07-21
Publication date: 2010-11-24
Anticipated expiration: 2030-07-21
Also published as: WO2011144136A1; CN101895373B

Abstract

The embodiment of the invention discloses a channel decoding method, system and device which are used in the technical field of communication. The method comprises the following steps: obtaining the channel decoding soft value of receiving signal frame; judging whether at least one voice coding parameter in the channel decoding soft value is needed to correct; if so, correspondingly correcting the determined voice coding parameter needed to be corrected, and using the corrected voice coding parameter as a channel decoding result to store and output. Therefore, after channel decoding and the correction of each voice coding parameter in the coding result, each voice coding parameter can correspond to the corresponding attribute feature and has little influence on the voice signal, thus increasing the accuracy of the decoding result.

Description

Channel decoding method, system and device

Technical field

The present invention relates to communication technical field, particularly channel decoding method, system and device.

Background technology

Generally speaking, the receiving terminal of current communications systems is independently when carrying out channel decoding and source encoding, the error rate of channel decoding is than higher like this, for example, at Qualcomm Code Excited Linear Prediction (QCELP) (Code-Excited Linear Predictive coding, CELP) the adaptive codebook index that obtains (adaptive codebook index, AI) parameter in most cases is gradual, if the AI parameter decoding value that channel decoding obtains and its previous decoding value relatively alter a great deal, the possibility of makeing mistakes so is just very big.And the information of message source and channel combined decoding method comprehensive channel and information source two aspects is made suitable decoding, has improved the accuracy of channel decoding.

Existing message source and channel interpretation method is predict the possibility of current decode results from historical decode results by certain Forecasting Methodology, comprises the value of bit value and direct results of predictive interpretation parameter in the results of predictive interpretation respectively.For the interpretation method of Prediction Parameters value, specifically realize as follows:

Carry out traditional channel decoding and obtain channel decoding soft value, calculate the metric of relevant parameter value according to channel decoding soft value; With the method for linear prediction, from each parameter metric of the current decode results of parameter prediction of historical decode results, the metric of last comprehensive twice acquisition obtains most probable decode results.

Above-mentioned when calculating the metric of relevant parameter value according to channel decoding soft value, be in the corresponding relation of bit soft information that stores and error probability thereof, search the metric of each bit in the channel decoding soft value, thereby obtain the metric of parameter, so only the metric according to this corresponding relation acquisition parameter value is very inaccurate.

Summary of the invention

The embodiment of the invention provides channel decoding method, system and device, can improve the accuracy of decode results.

A kind of channel decoding method of the embodiment of the invention comprises:

Obtain the channel decoding soft value of received signal frame;

The speech coding parameters that obtains current subframe from described channel decoding soft value is the value of declaring firmly;

If the received signal frame is correct speech frame, judge whether value of declaring firmly of at least one speech coding parameters in the current subframe satisfies the condition that presets, if do not satisfy, then definite described speech coding parameters need be revised;

To determining that the speech coding parameters that needs to revise carries out corresponding correction; Revised speech coding parameters is stored and exports as the channel decoding result.

The embodiment of the invention provides a kind of channel decoding method, comprising:

Obtain the channel decoding soft value of received signal frame;

If described received signal frame is not correct speech frame, determine that the speech coding parameters of received signal frame need be revised;

Speech coding parameters at least one received signal frame in the described channel decoding soft value carries out corresponding correction; Revised speech coding parameters is stored and exports as the channel decoding result.

The embodiment of the invention provides a kind of channel decoding system, comprising:

The first soft value acquiring unit is used to obtain the channel decoding soft value of received signal frame;

The value of declaring acquiring unit firmly, the speech coding parameters that is used for obtaining current subframe from described channel decoding soft value is the value of declaring firmly;

The first speech frame judging unit is used to judge whether described received signal frame is correct speech frame;

The condition judgment unit is used for when the described received signal frame of the described first speech frame judgment unit judges is correct speech frame, and whether the speech coding parameters of judging the current subframe that the described acquiring unit of the value of declaring firmly the obtains value of declaring firmly satisfies the condition that presets; If do not satisfy, then determine to need to revise described speech coding parameters;

First amending unit is used for described condition judgment unit is determined that the speech coding parameters that needs are revised carries out corresponding correction;

First stores output unit, is used for the revised speech coding parameters of described first amending unit is stored and exports as the channel decoding result.

A kind of channel decoding system that the embodiment of the invention provides comprises:

The second soft value acquiring unit is used to obtain the channel decoding soft value of received signal frame;

The second speech frame judging unit is used to judge whether described received signal frame is correct speech frame;

Second amending unit, being used for working as the described second speech frame judgment unit judges received signal frame is not correct speech frame, then the speech coding parameters at least one received signal frame in the described channel decoding soft value carries out corresponding correction;

Second stores output unit, is used for the revised speech coding parameters of described second amending unit is stored and exports as the channel decoding result.

A kind of receiving terminal that the embodiment of the invention provides comprises arbitrary channel decoding system that the invention described above embodiment provides.

As seen, the channel decoding method of the embodiment of the invention comprises: the channel decoding soft value of obtaining the received signal frame; Judge when at least one speech coding parameters need be revised in the channel decoding soft value, then need the speech coding parameters of correction to carry out corresponding correction, and revised speech coding parameters is stored and exports as the channel decoding result determining.Like this after channel decoding, through after revising, can both meet separately attributive character and also less to each speech coding parameters in the decode results, thereby improve the accuracy of decode results the voice signal influence.

Description of drawings

In order to be illustrated more clearly in the embodiment of the invention or technical scheme of the prior art, to do to introduce simply to the accompanying drawing of required use in embodiment or the description of the Prior Art below, apparently, accompanying drawing in describing below only is some embodiments of the present invention, for those of ordinary skills, under the prerequisite of not paying creative work, can also obtain other accompanying drawing according to these accompanying drawings.

Fig. 1 is the flow chart of a kind of channel decoding method of providing of the inventive method embodiment;

Fig. 2 be the high order bit that obtains speech coding parameters that the inventive method embodiment provides may value the flow chart of metric of correspondence;

Fig. 3 is the flow chart of the another kind of channel decoding method that provides of the inventive method embodiment;

Fig. 4 is based on the decoding architecture schematic diagram of the voice signal frame of EFR;

Fig. 5 is based in the decoding of voice signal frame of EFR the modification method flow chart to the FG parameter;

Fig. 6 is based in the decoding of voice signal frame of EFR the modification method flow chart to the AG parameter;

Fig. 7 is based in the decoding of voice signal frame of EFR the modification method flow chart to the AI parameter;

Fig. 8 is based in the decoding of voice signal frame of EFR the modification method flow chart to the LSF parameter;

Fig. 9 is the structural representation of the channel decoding system that provides of one embodiment of the invention;

Figure 10 is the structural representation of another channel decoding system of providing of system embodiment of the present invention;

Figure 11 is the structural representation of another channel decoding system of providing of system embodiment of the present invention.

Embodiment

Below in conjunction with the accompanying drawing in the embodiment of the invention, the technical scheme in the embodiment of the invention is clearly and completely described, obviously, described embodiment only is the present invention's part embodiment, rather than whole embodiment.Based on the embodiment among the present invention, those of ordinary skills belong to the scope of protection of the invention not making the every other embodiment that is obtained under the creative work prerequisite.

One embodiment of the invention provides a kind of channel decoding method, and flow chart can comprise as shown in Figure 1:

Step 101, obtain the channel decoding soft value of received signal frame;

The method of the embodiment of the invention can be applied in the voice signal decode procedure based on CELP, as enhanced full rate voice coding (Enhanced Full Rate speech codec, EFR) signal and adaptive multi-rate speech coding (Adaptive Multi Rate speech codec, AMR) decode procedure of signal etc.Also can be applied to based in other speech codings and the channel decoding process based on the video coding of other parameters.

Be understandable that, those skilled in the art will know that frame is to received signal carried out can obtaining decode results after the channel decoding, according to this decode results can be directly or this obtain channel decoding soft value indirectly, and the channel decoding soft value that obtains is the signal floating point values that comprises each speech coding parameters.Channel decoding soft value can directly be obtained from the result of channel decoding, for example: when the signal frame that receives based on EFR, demodulation bit soft information (DEM_SOFT_VAL) is input to after the channel decoding module deciphers, can decipher hard value, 189 channel decoding soft value and 378 demodulation bit external information by delivery channel, can directly from the result of channel decoding, obtain channel decoding soft value like this.

The speech coding parameters that comprises in the voice signal based on CELP can have constant codebook excitations (fixed codebook gain, FG), adaptive codebook excitation (adaptive codebook gain, AG), AI parameter and line spectrum parameter (line spectral frequencies, LSF).For example for an EFR signal frame, comprise 4 subframes in the channel decoding soft value of obtaining, the FG parameter of each subframe has 5 bits, and AG and AI parameter have 9 bits and 4 bits respectively, and in channel decoding soft value, have the LSF parameter of 38 bits, be divided into LSF1 to LSF5.

Step 102, speech coding parameters value of declaring firmly of from channel decoding soft value, obtaining current subframe;

Speech coding parameters in the channel decoding soft value can have a plurality of, can comprise the speech coding parameters of received signal and/or each speech coding parameters of each subframe.For example in the signal frame based on EFR, parameter F G, AG and AI are that each subframe is one group, and parameter L SF is each signal frame is one group, the speech coding parameters that can call the parameter that is similar to FG, AG and AI each subframe so can call the parameter that is similar to LSF the speech coding parameters of received signal.Present embodiment is to be applicable to the speech coding parameters of each subframe such as the channel decoding method of FG, AG and AI parameter.

And, the position of each speech coding parameters is fixed in the channel decoding soft value of signal frame, find out the channel decoding soft value of each speech coding parameters corresponding bit position in the current subframe earlier, is 0 with channel decoding soft value less than 1 bit position, is 1 with channel decoding soft value greater than 1 bit position, is speech coding parameters value of declaring firmly with regard to the bit value that has obtained each speech coding parameters like this.

Step 103, judge whether the received signal frame is correct speech frame, if, further execution in step 104 then;

Whether the signal frame that can judge reception in the following way is correct speech frame: if cyclic redundancy check (CRC) (the Cyclical Redundancy Check that definite frame to received signal carries out, CRC) verification is passed through, and the bit error rate of this signal frame correspondence (bit error rate, BER) value thinks then that less than the threshold value that presets the signal frame of this reception is correct speech frame; Otherwise think that the signal frame of this reception is not correct speech frame.

The signal frame that receives is carried out CRC check can judge that whether the received signal frame carries out verification correctly the time; Also can carry out verification, and check results is stored at the other times that receive signal frame.Like this CRC check of carrying out at the signal frame that judge to receive whether by the time, the direct check results that stores in the reading system, or carry out the CRC check of signal frame is earlier judged according to check results then.

In addition, not correct speech frame if judge the received signal frame, then can further carry out rapid 105;

104, judge at least one speech coding parameters in the current subframe firmly the value of declaring whether satisfy the condition that presets, if do not satisfy, then determine to need to revise speech coding parameters, execution in step 106; If satisfy, then do not need to revise, directly delivery channel is deciphered soft value;

Be appreciated that, prerequisite for the different speech coding parameter correspondence of current subframe is all different, here the condition that presets can be the attributive character value of each speech coding parameters, perhaps less condition value to the influence of voice signal, do not satisfy the condition that presets if judge, promptly do not meet the attributive character of speech coding parameters, or bigger to the voice signal influence, then determine to need to revise this speech coding parameters.

For example in the signal frame based on EFR, speech coding parameters can have different judgment modes for parameter F G, AG and the AI of parameter L SF and each subframe in the channel decoding soft value:

Particularly,, cause noise easily, then when judging, can judge the size of FG parameter,, then determine to need to revise the FG parameter if surpass certain value if when the FG parameter is big; For the AG parameter, can judge the difference etc. of the AG parameter of the AG parameter of current subframe and last subframe; For the AI parameter, whether the difference that can judge the inverse quantization value of this parameter in current subframe and the last subframe is in the AI scope that presets, if not in the AI scope that presets, then need to revise AI parameter or the like.

Be appreciated that the condition that presets correspondingly includes but not limited to following information here:

If speech coding parameters is the FG parameter, the condition that then presets comprises: the high order bit of the FG parameter of the current subframe value of declaring firmly is no more than a FG value that presets, and the FG scope that presetting of the FG parameter difference of the last subframe of the FG parameter of current subframe and storage; And/or the FG parameter of current subframe firmly the value of declaring be no more than the 2nd FG value that presets;

If speech coding parameters is the AG parameter, the condition that then presets comprises: the AG parameter of current subframe firmly the difference of the value of declaring and the AG parameter of the last subframe that stores less than an AG value that presets; And/or the AG parameter firmly the number of times that do not reduce continuously of the value of declaring less than the threshold value that presets;

If speech coding parameters is the AI parameter, the condition that then presets comprises: the difference of the inverse quantization value of the value of declaring is in the AI scope that presets firmly for this AI parameter in current subframe and last or preceding two subframes, and last signal frame is steady.Particularly, can whether judge less than a value that presets whether signal frame is steady by the difference of determining the minimum and maximum value of the AI parameter inverse quantization value of four subframes in the current demand signal frame, if less than, illustrate that then this signal frame is steady, otherwise illustrate that this signal frame is not steady.

105, obtain the metric of the possible value correspondence of high N position bit of speech coding parameters, with the speech coding parameters high order bit possibility value replacement of the high N position bit value usefulness metric minimum of the value of declaring firmly;

Here the high N position bit of speech coding parameters may value be speech coding parameters high N position bit 2 ^NIndividual possibility value, wherein N is less than or equal to the bit of speech coding parameters, as speech coding parameters 5 bits is arranged, and then its high 3 bits possibility value is 8 values between 000 to 111.Metric mainly is meant the transition probability of speech coding parameters and last subframe in the current subframe, reach transition probability with other speech coding parameters, and the bit error probability of this possibility value, and the bit correct probability equiprobability factor of this probable value and, specifically can be with reference to the description of subsequent embodiment.

The corresponding metric of each high N position bit possibility value, the high N position bit value of the value of declaring may the value replacement with the high N position bit of metric minimum firmly with speech coding parameters.

Step 105 back execution in step 104.

Step 106, the speech coding parameters that definite needs are revised carry out corresponding correction;

When revising, also can revise according to the attributive character of each speech coding parameters or to the influence of voice signal, the mode of revising has multiple, the speech coding parameters of attributive character will be adapted to the attributive character that meets separately as long as will not meet separately, perhaps will be adapted to less the getting final product of voice signal influence the bigger speech coding parameters of voice signal influence.

For example in the signal frame based on EFR, for the correcting mode of each parameter may for: reduce the FG parameter to the value that presets, or the value of the bit of channel decoding soft value absolute value minimum in the high order bit of excessive FG parameter be changed to 0 etc.; Excessive AG parameter is reduced to value that presets etc.; AI parameter with the current subframe of AI parameter substitution of last subframe.Step 106 back execution in step 107.

Step 107, revised speech coding parameters is stored and exports as the channel decoding result.

The channel decoding method of the embodiment of the invention comprises: the channel decoding soft value of obtaining the received signal frame; The speech coding parameters that obtains current subframe from channel decoding soft value is the value of declaring firmly; If the received signal frame is correct speech frame, judge at least one speech coding parameters in the current subframe firmly the value of declaring whether satisfy the condition that presets, if do not satisfy, then need the speech coding parameters of correction to carry out corresponding correction, and revised speech coding parameters is stored and exports as the channel decoding result determining.Like this after channel decoding, the speech coding parameters of the needs correction in the decode results through after revising, is made speech coding parameters can both meet separately attributive character or also less to the voice signal influence, thereby can improve the accuracy of decode results.

Be appreciated that in a specific embodiment, when carrying out the metric that the high N position bit that obtains speech coding parameters in the above-mentioned steps 105 may the value correspondence, can realize as follows, flow chart as shown in Figure 2:

A1, obtain with the corresponding high N of speech coding parameters position than the ultrahard value of declaring;

Can obtain high N position than the ultrahard value of declaring by extracting the value of declaring firmly from speech coding parameters, wherein N is less than or equal to the speech coding parameters number of bits of the value of declaring firmly.

The channel metric of B1, the high N of acquisition position bit possibility value:

The value of high N position than corresponding bits position in ultrahard value of declaring and the high N position bit possibility value compared, obtain the identical or different comparative result of value of bit; According to this comparative result, and first corresponding relation of bit channel information that presets and metric, obtain the channel metric of each bit, and with the channel metric sum of each bit as this high N position bit may the value correspondence the channel metric; Wherein the bit channel information comprises the combination of channel decoding soft value and following any or a plurality of information of bit: average BER value, the speed sign as when identifying full rate still be the sign etc. of half rate;

Be appreciated that in first corresponding relation and comprise: the corresponding relation of the corresponding relation of bit channel information and bit correct probability and bit channel information and bit error probability.If the comparative result that obtains is that the value of bit is inequality,, obtain the channel metric of a bit error probability as this bit then according to the corresponding relation of bit channel information in first corresponding relation and bit error probability; If comparative result is that the value of bit is identical,, obtain the channel metric of a bit correct probability as this bit then according to the corresponding relation of bit channel information in first corresponding relation and bit correct probability.

Wherein the sign of the speed in the channel information can be from channel decoding soft value, and the speed flag bit of this subframe obtains; And average BER is meant the BER of current channel decoding soft value correspondence, directly receives to obtain.

For example: high 4 that suppose to obtain in the steps A 2 is 1001 than the ultrahard value of declaring, get high 4 bits possibility value 0101 in high 4 bits possibility value 0000 to 1111, with the value of declaring firmly with may value in the value of corresponding bits position compare, the comparative result that obtains is: 3rd, 4 value difference, and the 1st, 2 value is identical; For the 3rd, 4 channel metric, obtain the 3rd, 4 bit channel informations earlier, and in the corresponding relation of bit channel information and bit error probability, search, obtain two bit error probabilities respectively as the 3rd, 4 channel metric; For the 1st, 2 channel metric, obtain the 1st, 2 bit channel informations earlier, and in the corresponding relation of bit channel information and bit correct probability, search, obtain two bit correct probabilities respectively as the 1st, 2 channel metric; 1 to 4 channel metric addition is obtained the channel metric of this high 4 bits possibility value 0101.

The information source metric of C1, the high N of acquisition position bit possibility value:

According to second corresponding relation of information source information that presets and n rank probability, obtain the information source metric of a n rank probability as this high N position bit possibility value correspondence; Wherein: n is more than or equal to 0; Information source information comprises that high N position bit may value and the combination of following any or a plurality of information: the high N position of the speech coding parameters of last subframe is than the ultrahard value of declaring, the high N position of other speech coding parameters of last subframe is than the ultrahard value of declaring, the high N position of this speech coding parameters of other speed is than the ultrahard value of declaring, the speech coding parameters of last subframe is transformed into high N position with the speech coding parameters same rate of current subframe than the ultrahard value of declaring, the correlation of current subframe spectral line and last subframe spectral line etc.

Be appreciated that second corresponding relation can comprise (following is that transition probability is the example explanation with 1 rank probability):

1, the speech coding parameters of last subframe is to the transition probability of current this speech coding parameters of subframe, when obtaining the information source metric, consider the correlation of same speech coding parameters in two continuous subframes like this, as mtr_fg[fg_pre] [fg_cur], FG parameter high order bit value of declaring firmly of the last subframe of [fg_pre] expression wherein, the high N position of the current subframe FG parameter of [fg_cur] expression is than the ultrahard value of declaring.

2, other speech coding parameters of last subframe are to the transition probability of current this speech coding parameters of subframe, when obtaining the information source metric, consider the correlation of different speech coding parameter in two continuous subframes like this, as ai_ag[ai_pre] [ag_cur], FG parameter high order bit value of declaring firmly of the last subframe of [ai_pre] expression wherein, the transition probability of the AI parameter of the last subframe of value representation to the AG parameter of current subframe declared in the high N position of the current subframe AG parameter of [ag_cur] expression than ultrahard.

3, second corresponding relation can also comprise mtr_fg[lsf_state] [fg_pre] [fg_cur], wherein [lsf_state] is meant the correlation of current subframe spectral line and last subframe spectral line, make the information source metric more can reflect the actual conditions of the voice signal frame of current subframe, judge this N high order bit possibility value and the high order bit gap between the value of declaring firmly more accurately.

4, for voice signal frame based on AMR, it is variable bit rate, then second corresponding relation can also comprise: the transition probability between this speech coding parameters of other speed and the same speech coding parameters of current subframe, the high N position bit that needs to obtain earlier the speech coding parameters of current subframe may value, and in the system the high N position of that store and same speech coding parameters current subframe different rates than the ultrahard value of declaring; In second corresponding relation, search corresponding transition probability again.

If the speed difference of continuous two subframes in 5 the signal frames based on AMR, second corresponding relation can also comprise: the transition probability of last subframe when current subframe is carried out switching rate, particularly, the first speech coding parameters with last subframe of needs is transformed into the speech coding parameters same rate with current subframe, but high N position, conversion back is than the ultrahard value of declaring, obtain the high N position bit possibility value of the speech coding parameters of current subframe again, and in second corresponding relation, find corresponding transition probability.

When second corresponding relation comprises above-mentioned 1 to 5 any or a plurality of corresponding relation, can obtain one or more transition probabilities earlier, and the information source metric can be a transition probability that obtains, or a plurality of transition probability sum.

Need to prove that second corresponding relation can be divided into the corresponding relation of odd number subframe information source information and n rank probability and the corresponding relation of even number subframe information source information and n rank probability.When speech coding parameters comprises AI, the redundancy of AI parameter is bigger in continuous two subframes, be that the AI parameter changes not quite, so the AI parameter of even number subframe adopts absolute the quantification when coded quantization, the AI parameter of odd number subframe adopts relatively and quantizes, then when the information source metric of the high order bit possibility value correspondence that obtains the AI parameter:

Judge that earlier current subframe is even number subframe or odd number subframe, if the odd number subframe then according to the corresponding relation of odd number subframe information source information in second corresponding relation and n rank probability, obtains a n rank probability; If the even number subframe then according to the corresponding relation of even number subframe information source information in second corresponding relation and n rank probability, obtains a n rank probability.

D1, high N position bit that step B1 is obtained may value the high N position bit that obtains of channel metric and step C1 may value the information source metric and as the corresponding metric of the possible value of this high N position bit.Be appreciated that step B1 and C1 do not have absolute ordinal relation, can carry out simultaneously, also can successively carry out.

By the cycling of above-mentioned steps A1, can obtain the speech coding parameters metric that 2N high N position bit may the value correspondence in the value of declaring firmly respectively to D1.

Another embodiment of the present invention also provides a kind of channel decoding method, and flow chart can comprise as shown in Figure 3:

A2, obtain the channel decoding soft value of received signal frame;

When obtaining channel decoding soft value, can directly from the result of channel decoding, extract the channel decoding soft value of received signal frame; Also can extract the external information of demodulation bit from the result of channel decoding, the relation according between coding multinomial and the coding output sequence reaches the external information of demodulation bit and the relation between the demodulation bit soft information, obtains the channel decoding soft value of received signal frame.

Concrete preparation method described in the embodiment, does not repeat them here as shown in Figure 1.

B2, judge whether the received signal frame is correct speech frame, if not, determine that then the speech coding of received signal frame need be revised execution in step C2; If, then do not need to revise, directly delivery channel is deciphered soft value;

Be appreciated that, when whether needing to revise, can judge whether the received signal frame is correct speech frame at the speech coding parameters of judging the received signal frame (as LSF parameter etc.), if not, the speech coding parameters of then determining the received signal frame needs to revise, then execution in step C2.

Judging whether when being correct speech frame, pass through if determine the CRC check that frame to received signal carries out, and the BER value of this signal frame correspondence is thinking then that less than the threshold value that presets the signal frame of this reception is correct speech frame; Otherwise think that the signal frame of this reception is not correct speech frame.

C2, the speech coding parameters of at least one received signal frame in the channel decoding soft value is carried out corresponding correction;

Generally speaking, the position of the speech coding parameters of received signal frame in channel decoding soft value is changeless, can behind the acquisition speech coding parameters, revise in channel decoding soft value earlier again; Such as, the LSF parameter is one group with a signal frame, be the speech coding parameters of received signal frame, can revise according to the attributive character of each speech coding parameters or to the influence of voice signal, the mode of revising has multiple, the speech coding parameters of attributive character will be adapted to the attributive character that meets separately as long as will not meet separately, perhaps will be adapted to less the getting final product of voice signal influence the bigger speech coding parameters of voice signal influence.

Such as, when the LSF parameter is revised: can the line spectrum parameter L SF in the channel decoding soft value smoothly be calculated with the state parameter of the last signal frame of current storage, with the correction value of the value after smoothly calculating as line spectrum parameter L SF.

D2, revised speech coding parameters is stored and exports as the channel decoding result.

As seen, the channel decoding method among the inventive method embodiment comprises: the channel decoding soft value of obtaining the received signal frame; If the received signal frame is not correct speech frame, at least one speech coding parameters in the channel decoding soft value is carried out corresponding correction, and revised speech coding parameters is stored and exports as the channel decoding result.Like this after channel decoding, to the speech coding parameters of the received signal frame of the needs correction in the decode results through after revising, make speech coding parameters can both meet separately attributive character or also less, thereby can improve the accuracy of decode results voice signal influence.

In the above-mentioned channel decoding method embodiment illustrated in fig. 1 and embodiment illustrated in fig. 3, need all to judge whether at least one speech coding parameters needs to revise, revise if desired, then revise, and, can use the information (being information source information) of the speech coding parameters of the last subframe of current storage or last signal frame in the process of process of judge revising or correction for a speech coding parameters.Different is be the processing for the speech coding parameters of a subframe in the received signal frame in the method embodiment illustrated in fig. 1, and method embodiment illustrated in fig. 3 to be the processing for the speech coding parameters of received signal frame.

Below so that the channel decoding method of the embodiment of the invention to be described based on the voice signal frame of EFR, the structure of deciphering at receiving terminal comprises as shown in Figure 4: channel decoding module, parameter Estimation module and source coding module, particularly:

When communication ends receives signal frame based on EFR, with 378 demodulation bit soft information (demodulated soft vale, DEM_SOFT_VAL) input channel decoding module, the then processing deciphered of channel decoding module, can give the hard value of parameter Estimation module delivery channel decoding, 189 channel decoding soft value (decoded soft vale, DEC_SOFT_VAL) and/or 378 demodulation bit external information (extrinsic soft value, EXT_SOFT_VALUE);

Communication ends will receive based on silence description frames in the signal frame of EFR (Silence Descriptor, SID), speech frame descriptor frame (Speech Descriptor, SP), the sign of CRC check, (Signalto Noise Ratio SNR) is input to the parameter Estimation module with the BER equivalence to signal to noise ratio;

The parameter Estimation module is receiving above-mentioned these information, obtain channel decoding soft value earlier, wherein speech coding parameters carried out the judgement that whether will revise, revise speech coding parameters if desired, then carry out the processing of above-mentioned these steps of process such as correction of corresponding speech coding parameters, value of declaring firmly (HARD_VAL) to source coding module output vocoder frames is revised decode results, has finished the judgement of bad frame simultaneously, output bad frame identity (BFI).

And the processing method of parameter Estimation module is the described channel decoding method of the embodiment of the invention in the said structure, particularly, realizes as follows:

1, the parameter Estimation module is obtained channel decoding soft value:

If the output of channel decoding module comprises channel decoding soft value, then the parameter Estimation module can directly directly be obtained from the output of channel decoding module; If the output of channel decoding module includes only 378 demodulation bit external information, then can be by the relation between coding multinomial and the coding output sequence, reach the external information of demodulation bit and the relation between the demodulation bit soft information, obtain the channel decoding soft value of received signal frame.

At the coding side of communication, note coding multinomial is:

[\begin{matrix} c_{00}, c_{01}, . . . . . . . . . . ., c_{0, K - 1} \\ c_{10}, c_{11}, . . . . . . . . . . . ., c_{1, K - 1} \\ . . . . . . . . . . . . . . . . . . . . . . . . . . . . . \\ c_{k - 1,0}, c_{k - 1,1}, . . ., c_{k - 1, K - 1} \end{matrix}]

Wherein every row is represented a coding multinomial, and note coding input bit sequence is u in addition ₀, u ₁, u ₂..., the coding output sequence is v ₀, v ₁, v ₂..., then the cataloged procedure of coding side is: c _I0u _n+ c _I1u _N-1+ ...+c _{I, K-1}u _N-K+1=v _Kn+i, wherein: i=0,1 ..., k-1, K are code length.

Obtain by the relation between above-mentioned coding multinomial and the coding output sequence: u _n=f (v ⁽⁰⁾, v ⁽¹⁾..., v ^(t)) n=0,1,2..., wherein t and f are by coding multinomial and cataloged procedure decision.Have generally speaking: u _n=v ⁽⁰⁾+ v ⁽¹⁾+ ...+, v ^(t)N=0,1,2....

For example, in EFR, the coding multinomial is Then cataloged procedure is

u _n+u _n-3+u _n-4＝v _2n

u _n+u _n-1+u _n-3+u _n-4＝v _2n+1 (1)

Can obtain by two equatioies in above-mentioned (1): u _N-1=v _2n+ v _2n+1, be equivalent to u _n=v _{2 (n+1)}+ v _{2 (n+1)+1}In this example, t equals 1, v ⁽⁰⁾, v ⁽¹⁾Be respectively v ^{2 (n+1)}, v _{2 (n+1)+1}

Then at the decoding end, note demodulation bit soft information is demllr (i), i=0 ..., 377, demodulation bit external information is demextr (i), i=0 ..., 377, the soft information of demodulation coding that then has the decoding gain is demllrextr (v ⁽ⁱ⁾)=demllr (v ⁽ⁱ⁾)+demextr (v ⁽ⁱ⁾), i=0 wherein, 1 ..., t, and the parameter Estimation module obtains bit u _nThe estimated value of channel decoding soft value be:

decllr(u _n)＝(-1) ^*sign(v ⁽⁰⁾) ^*sign(v ⁽¹⁾) ^*... ^*sign(v ^(t))

^*min(abs(demllrextr(v ⁽⁰⁾))，abs(demllrextr(v ⁽¹⁾))，...，abs(demllrextr(v ^(t))))n＝0，1，2，......

Then in the decoding based on EFR, the channel decoding soft value that obtains the received signal frame is:

decllr(u _n)＝(-1) ^*sign(demllrextr(v _2(n+1)) ^*sign(demllrextr(v _2(n+1)+1))

^*min(abs(demllrextr(v _2(n+1)))，abs(demllrextr(v _2(n+1)+1)))，i＝0，...，188

Wherein min represents to get minimum value; Abs represents to take absolute value; Sign is a piecewise function, and the symbol of certain number is got in expression, and for example, the symbol value of the number more than or equal to 0 is 1, and the symbol value of the number less than 0 is-1.

Can obtain 189 channel decoding soft value at last, and last 4 be the tail bit, obtain 185 channel decoding soft value after casting out.

2, the parameter Estimation module is carried out the correction of speech coding parameters:

Obtained channel decoding soft value by said method parameter Estimation module based on EFR voice signal frame, in this channel decoding soft value, comprise 4 subframes, have 38 bits comprising the LSF parameter, be divided into LSF1～LSF5, and the FG parameter of each subframe has 5 bits, and AI, AG parameter have 9 bits and 4 bits respectively.Below describe in detail respectively:

(1) for the correction of FG parameter, can realize as follows, flow chart as shown in Figure 5:

201, parameter Estimation module 5 bit channels that extract earlier the FG parameter corresponding bit position of current subframe from channel decoding soft value are deciphered soft value, thus the FG parameter that obtains 5 bits in the current subframe according to this channel decoding soft value value of declaring firmly.

202, the parameter Estimation module judges according to the SP that receives whether the current demand signal frame is speech frame, if then execution in step 203; If not, then execution in step 204.

203, judge whether the current demand signal frame is correct speech frame, if then execution in step 209, if not, then execution in step 205.

Particularly, judge according to the sign of the CRC check that receives whether the CRC check of current demand signal frame is passed through, and judge that according to the BER signal that receives whether this BER value is less than the threshold value that presets, if CRC check by and the BER value less than the threshold value that presets, determine that then the current demand signal frame is correct speech frame.

204, preserve the FG parameter of current demand signal frame the value of declaring high order bit be 0 firmly after, execution in step 213.

205, the FG parameter of getting 5 bits is firmly in the value of declaring, high 4 than the ultrahard value of declaring, be assumed to be 0110.

Need to prove, be that example describes to get high 4 than the ultrahard value of declaring in the present embodiment, and for the situation of the high N of getting of other position than the ultrahard value of declaring, processing method similarly.Step 205 back execution in step 206.

206, obtain the metric of a high order bit possibility value correspondence:

Be appreciated that since got in the FG parameter high 4 than the ultrahard value of declaring, then high order bit may value be arbitrary value of 16 high order bit values in 0000～1111, supposing to get a high order bit probable value is 1010.

Obtain corresponding channel metric earlier: with 1010 with the value of declaring 0110 comparison firmly, the value that obtains the 1st, 2 is identical, the comparative result that the 3rd, 4 value is different; The channel decoding soft value of then using these 3,4 bits is as index, in the channel decoding soft value and bit error probability correspondence table of the bit that stores, obtains two bit error probabilities respectively as the channel metric of the 3rd, 4 bits; As index, in the channel decoding soft value and bit correct probability correspondence table of the bit that stores, obtain two bit correct probabilities with the channel decoding soft value of these 1,2 bits respectively as the channel metric of the 1st, 2 bits; And with the channel decoding metric addition of 4 bits, obtain may value 1010 correspondences the channel metric.

Obtain corresponding information source metric again: with the high order bit of the last subframe that stores firmly the value of declaring be index with the possible value of this high order bit, the transition probability that finds a correspondence in the transition probability table is as the information source metric.

With above-mentioned information source metric that obtains and the addition of channel metric, obtain the metric of possibility value 1010 correspondences.Step 206 back carries out 207.

207, judge whether the metric that 16 high order bits may the value correspondences has all calculated, if, then after execution in step 208, execution in step 210; If not, then return the metric that obtains another high order bit possibility value correspondence in the execution in step 206.

208, get 16 high order bits may the metric of value correspondences in the high order bit of minimum metric may value, as the high order bit of the FG parameter value of declaring firmly.

209, determine from BER and the SNR that receives whether current subframe is reliable, in certain adjustable extent, if reliable, then execution in step 212 for BER that promptly receives and SNR; If unreliable, then execution in step 210.

210, the FG parameter of judging current subframe firmly the value of declaring whether satisfy the condition that presets, if satisfy, then execution in step 212, if do not satisfy, then execution in step 211.

Here the condition that presets includes but not limited to: the FG parameter of the current subframe high order bit of the value of declaring firmly is no more than a FG value that presets, and the FG scope that presetting of the FG parameter difference of the last subframe of the FG parameter of this subframe and storage; And/or the FG parameter of current subframe firmly the value of declaring be no more than the 2nd FG value that presets etc.

211, carry out the correction of FG parameter: reduce the FG parameter, can be with the high order bit of FG parameter in the soft value minimum bit soft value be changed to 0 etc.;

212, preserve to revise after the parameter high order bit of the value of declaring firmly of this subframe FG.Step 212 back execution in step 213.

213, the FG parameter of this subframe value of declaring firmly after output is revised.Step 213 back execution in step 214.

214, whether the correction of judging the FG parameter of 4 subframes in the channel decoding soft value is handled, if, process ends then, if not then returning execution in step 201, the FG parameter of obtaining next subframe is the value of declaring firmly.

(2) for the correction of AG parameter, can realize as follows, flow chart as shown in Figure 6:

301, parameter Estimation module 4 bit channels that extract earlier the AG parameter corresponding bit position of a subframe from channel decoding soft value are deciphered soft value, thus the AG parameter that obtains 4 bits in the current subframe according to this channel decoding soft value value of declaring firmly.

302, judge according to the SP that receives whether the current demand signal frame is speech frame, if then execution in step 303; If not, then execution in step 313.

303, judge whether the current demand signal frame is correct speech frame, if then execution in step 308, if not, then execution in step 304.

304, the AG parameter of getting 4 bits is firmly in the value of declaring, high 3 than the ultrahard value of declaring, be assumed to be 010.Step 304 back execution in step 305.

305, obtain the metric of a high order bit possibility value correspondence: step 305 back execution in step 306.

Be appreciated that, since got in the AG parameter high 3 than the ultrahard value of declaring, then high order bit may value be 000～111 arbitrary value in the high order bit possibility value, supposing to get a high order bit may value be 110, and obtain the channel metric road decoding metric and the information source metric of 110 correspondences, and addition obtain may value 110 correspondences metric.The described mode of step 206 is similar in concrete acquisition methods and the FG parameter, does not repeat them here.

306, whether the metric of judging 8 high order bits possibility value correspondences has all calculated, if then execution in step 307; If not, then return the metric that obtains another high order bit possibility value correspondence in the execution in step 305.

307, get 8 high order bits may the metric of value correspondences in the high order bit of minimum metric may value, as the high order bit of the AG parameter value of declaring firmly.Step 307 back execution in step 308.

308, the AG parameter of judging this subframe firmly the value of declaring whether satisfy the condition that presets, if satisfy, then execution in step 310, if do not satisfy, then execution in step 309.

Here the condition that presets comprises: the AG parameter of current subframe firmly the difference of the value of declaring and the AG parameter of the last subframe that stores less than an AG value that presets; And/or the AG parameter firmly the number of times that do not reduce continuously of the value of declaring less than the 2nd AG value that presets etc.

309, carry out the correction of AG parameter: reduce the AG parameter to meeting the AG attributive character, and to the less value of voice signal influence.Step 309 back execution in step 310.

310, preserve to revise after the parameter high order bit of the value of declaring firmly of this subframe AG.Step 310 back execution in step 311.

311, the AG parameter of this subframe value of declaring firmly after output is revised.Step 311 back execution in step 312.

312, whether the correction of judging the AG parameter of 4 subframes in the channel decoding soft value is handled, if, process ends then, if not then returning execution in step 301, the AG parameter of obtaining next subframe is the value of declaring firmly.

313, preserve the AG parameter of current demand signal frame the value of declaring high order bit be 0 firmly after, execution in step 311.

(3) for the correction of AI parameter, can realize as follows, flow chart as shown in Figure 7:

401, parameter Estimation module 9 bit channels that extract earlier the AI parameter corresponding bit position of a subframe from channel decoding soft value are deciphered soft value, thus the AI parameter that obtains 9 bits in the current subframe according to this channel decoding soft value value of declaring firmly.

402, judge according to the SP that receives whether the current demand signal frame is speech frame, if then execution in step 403; If not, then execution in step 417.

403, judge whether the current demand signal frame is correct speech frame, if then execution in step 408, if not, then execution in step 404.

404, the AI parameter of getting 9 bits is firmly in the value of declaring, high 4 than the ultrahard value of declaring, be assumed to be 0010.Step 404 back execution in step 405.

405, obtain the metric of a high order bit possibility value correspondence: step 405 back execution in step 406.

Be appreciated that, since got in the AI parameter high 4 than the ultrahard value of declaring, then high order bit possibility value is the arbitrary value in 0000～1111, supposing to get a high order bit may value be 0110, and obtain the channel metric road decoding metric and the information source metric of 0110 correspondence, and addition obtain may value 0110 correspondence metric.

The acquisition methods of concrete channel metric does not repeat them here as described in the step 206 in the FG parameter; And when obtaining the information source metric, need will judge that current subframe is even number subframe or odd number subframe, and according to the even number subframe information source information that stores and the corresponding relation of transition probability, or the corresponding relation of odd number subframe and transition probability obtains corresponding information source metric.

406, whether the metric of judging 16 high order bits possibility value correspondences has all calculated, if, then in execution in step 407; If not, then return the metric that obtains another high order bit possibility value correspondence in the execution in step 405.

407, get 8 high order bits may the metric of value correspondences in the high order bit of minimum metric may value, as the high order bit of the AI parameter value of declaring firmly.Execution in step 408 after the step 407

408, the AI parameter of judging this subframe firmly the value of declaring whether satisfy the condition that presets, if satisfy, then execution in step 410, if do not satisfy, then execution in step 409.

Here the condition that presets comprises: the difference of the inverse quantization value of this AI parameter is in the AI scope that presets in the value of declaring and the last subframe firmly for current subframe, and last signal frame is steady etc.Obtaining last signal frame during information whether stably, whether stably last signal frame that can be by reading storage sign obtains.

409, carry out the correction of AI parameter: with the AI parameter of the current subframe of AI parameter substitution of last subframe.Step 409 back execution in step 410.

410, calculate and preserve revise after the high order bit inverse quantization value of the value of declaring firmly of this subframe AI parameter.Step 410 back execution in step 411.

411, judge that current subframe is even number subframe or odd number subframe, if the odd number subframe, then execution in step 412, if the even number subframe, then execution in step 413.

412, the high order bit of not preserving AI parameter in current subframe value of declaring firmly, and revise in the current subframe AI parameter low 2 bit values of the value of declaring firmly.Step 412 back execution in step 414.

413, the high order bit of preserving AI parameter in current subframe value of declaring firmly.Step 413 back execution in step 414.

414, the AI parameter of this subframe value of declaring firmly after output is revised.Step 414 back execution in step 415.

415, whether the correction of judging the AI parameter of 4 subframes in the channel decoding soft value is handled, if then execution in step 416, if not then returning execution in step 401, the AI parameter of obtaining next subframe is the value of declaring firmly.

416, judge whether current signal frame based on EFR is steady, and carry out mark according to judged result, and process ends.

Particularly, whether the difference of minimum and maximum value of AI parameter inverse quantization value of determining four subframes in the current demand signal frame less than a value that presets, if less than, illustrate that then this signal frame is steady, otherwise not steady., can whether steadily carry out mark to the current demand signal frame in relevant information herein, the mode of mark can be with reference to the mode of prior art, as long as whether steadily can express the current demand signal frame.

417, with the AI parameter of current demand signal frame firmly the value of declaring and the inverse quantization value be changed to 0 after, and execution in step 418.

418, with the current demand signal frame flag for not steady, and process ends.

(4) for the correction of LSF parameter, can realize as follows, flow chart as shown in Figure 8:

501, the parameter Estimation module obtains the state parameter of front signal frame.

Here state parameter can be the parameters such as inverse quantization value of the LSF parameter that comprises the front signal frame.

502, judge according to the SP that receives whether the current demand signal frame is speech frame, if then execution in step 503; If not, process ends then.

503, judge to the current demand signal frame whether be correct speech frame, if, then in execution in step 504 back execution in step 507, if not, then execution in step 505.

Particularly, judge according to the CRC signal that receives whether the CRC check of current demand signal frame is passed through, and judge that according to the BER signal that receives whether this BER value is less than the threshold value that presets, if CRC check by and the BER value less than the threshold value that presets, determine that then the current demand signal frame is correct speech frame.

504, obtain the inverse quantization value of LSF parameter in the current demand signal frame, and directly export the LSF parameter of current demand signal frame.Step 504 back execution in step 507.

When the inverse quantization value of the LSF parameter of obtaining the current demand signal frame, extract the channel decoding soft value of LSF parameter corresponding bit position in the first channel decoding soft value, obtain LSF parameter value of declaring firmly according to this channel decoding soft value, and carry out inverse quantization.

505, the LSF parameter is revised: according to the state parameter that obtains in the step 501 the LSF parameter of current demand signal frame is carried out smoothly, and the value after will smoothly calculating is as the correction value of LSF parameter.Step 505 back execution in

step

506 and 507.

Particularly, earlier the LSF parameter of current demand signal frame is carried out inverse quantization, and the LSF parameter inverse quantization value of state parameter and current subframe is smoothly calculated, at last with the LSF parameter quantification after level and smooth.

506, with revised LSF parameter output.

507, upgrade according to the state parameter of the revised LSF parameter of step 505, or upgrade the state parameter of current storage according to the LSF parameter of output in the step 504 current storage.

Above-mentioned in the decode procedure based on the EFR signal frame, parameter Estimation obtains the channel decoding soft value of received signal frame earlier, and judge whether speech coding parameters wherein needs to revise, if desired, then revise accordingly, the speech coding parameters that obtains after revising like this can both meet separately attributive character and also less to the voice signal influence, thereby has improved the accuracy of decode results.

The method of the embodiment of the invention can also be applied to the decode procedure based on the AMR signal frame, and concrete method is similar to above-mentioned decode procedure based on the EFR signal frame.But because AMR is the signal frame of variable bit rate, the speed of continuous two signal frames may be different, when the metric of the high order bit possibility value correspondence that obtains speech coding parameters: for the channel metric, can also further consider the speed of current subframe, the speed sign that promptly can comprise the channel decoding soft value of bit and current subframe in first corresponding relation is with two couple's one of metric corresponding relation; For the information source metric, can consider the transition probability of speech coding parameters between the different rates, or the 0 rank probability distribution that adopts speech coding parameters is as metric etc.

And the speech coding frame format that based on speed in the AMR signal frame is 12.2k is with the same based on the encoded frame format of EFR signal frame, then can be described with reference to above-mentioned interpretation method based on the EFR signal frame based on the interpretation method of this speed in the AMR signal frame, do not repeat them here.

But the encoded frame format of other speed is then different, mainly be different with the interpretation method of FG parameter for AG, based on removal rates in the AMR signal frame is the speech coding of 12.2k and 7.95k, the speech coding of other speed all adopts the quantification of uniting of AG and FG parameter, i.e. index of record in vocoder frames.Then when decoding, can table look-up by this index earlier and obtain the inverse quantization value of AG and FG parameter respectively, thereby obtain AG and FG value of declaring firmly separately; And the judgement of whether revising according to value of declaring firmly of each autoregressive parameter, and when the needs correction, revise accordingly.

One embodiment of the invention also provides a kind of message source and channel decoding system, and its structure can comprise as shown in Figure 9:

The first soft value acquiring unit 10 is used to obtain the channel decoding soft value of received signal frame;

The first soft value acquiring unit 10 can directly extract the channel decoding soft value of received signal frame from the result of channel decoding when obtaining channel decoding soft value; Also can extract the external information of demodulation bit from the result of channel decoding, the relation according between coding multinomial and the coding output sequence reaches the external information of demodulation bit and the relation between the demodulation bit soft information, obtains the channel decoding soft value of received signal frame.

The value of declaring acquiring unit 120 firmly, the speech coding parameters that is used for obtaining current subframe from the channel decoding soft value that the first soft value acquiring unit 10 obtains is the value of declaring firmly, and speech coding parameters can comprise any or a plurality of following information: constant codebook excitations FG, adaptive codebook excitation AG and adaptive codebook index AI;

The position of each speech coding parameters is fixed in the channel decoding soft value of signal frame, the value of declaring acquiring unit 120 is found out the channel decoding soft value of each speech coding parameters corresponding bit position in the current subframe earlier firmly, is 0 with channel decoding soft value less than 1 bit position, is 1 with channel decoding soft value greater than 1 bit position, is speech coding parameters value of declaring firmly with regard to the bit value that has obtained each speech coding parameters like this.

The first speech frame judging unit 220 is used to judge whether the received signal frame is correct speech frame, if then condition judgment unit 320 is further judged;

When whether the first speech frame judging unit 220 is correct speech frame at the signal frame that judge to receive, concrete is if the CRC check that definite frame to received signal carries out is passed through, and the BER value of this signal frame correspondence thinks then that less than the threshold value that presets this signal frame is correct speech frame; Otherwise just be not.

Condition judgment unit 320, whether the speech coding parameters that is used to judge the current subframe that the value of declaring acquiring unit 120 firmly the obtains value of declaring firmly satisfies the condition that presets; If do not satisfy, then determine to need to revise speech coding parameters;

Here the condition that presets includes but not limited to following information:

If speech coding parameters is FG, the condition that presets can comprise: the high order bit of the constant codebook excitations FG of the current subframe value of declaring firmly is no more than a FG value that presets, and the constant codebook excitations FG difference of the last subframe of the constant codebook excitations FG of current subframe and storage is no more than the FG scope that presets; And/or the constant codebook excitations FG of current subframe firmly the value of declaring less than the 2nd FG value that presets;

If speech coding parameters is AG, the condition that presets can comprise: the adaptive codebook of current subframe excitation AG firmly the difference of the value of declaring and the adaptive codebook excitation AG of the last subframe that stores less than an AG value that presets, and/or adaptive codebook encourage AG firmly the continuous number of times that does not reduce of the value of declaring less than the threshold value that presets;

If speech coding parameters is AI, the condition that then presets comprises: the difference of the inverse quantization value of the value of declaring is in the AI scope that presets firmly for adaptive codebook index AI in current subframe and last or two subframes, and last signal frame is steady.

When definite speech coding parameters comprises FG and AI, the condition that then presets just comprises the condition about FG and AI parameter.

First amending unit 30 is used for then condition judgment unit 320 being determined that the speech coding parameters that need to revise carry out corresponding correction when condition judgment unit 320 is judged the speech coding parameters of current subframe the value of declaring is not satisfied the condition that presets firmly;

First amending unit 30 is when revising, also be according to the attributive character of each speech coding parameters and the influence of voice signal is revised, the speech coding parameters of attributive character is adapted to the attributive character that meets separately with not meeting separately, will be adapted to the voice signal influence less to the bigger speech coding parameters of voice signal influence.

For example in the signal frame based on EFR: reduce the FG parameter to the value that presets, or the value of the bit of channel decoding soft value absolute value minimum in the high order bit of excessive FG parameter is changed to 0 etc.; Excessive AG parameter is reduced to value that presets etc.; AI parameter with the current subframe of AI parameter substitution of last subframe.

First stores output unit 40, is used for first amending unit, 30 revised speech coding parameters are stored and export as the channel decoding result.

In the message source and channel decoding system of the embodiment of the invention: the first soft value acquiring unit 10 obtains the channel decoding soft value of received signal frame; The value of declaring acquiring unit 120 obtains the speech coding parameters of current subframe from channel decoding soft value firmly, when determining that by the first speech frame judging unit 220 and condition judgment unit 320 at least one speech coding parameters need be revised in the current subframe, then 30 pairs of first amending units determine that the speech coding parameters that needs to revise carries out corresponding correction, store output unit 40 by first revised speech coding parameters is stored and exports as the channel decoding result.Like this after channel decoding, through after revising, can both meet separately attributive character and also less to each speech coding parameters in the decode results, thereby improve the accuracy of decode results the voice signal influence.

In a specific embodiment, with reference to shown in Figure 10, can also comprise in the channel decoding system of present embodiment: substituting unit 420, be used for when the first speech frame judging unit 220 judges that the received signal frames are not correct speech frames, obtain the metric of the possible value correspondence of high N position bit of speech coding parameters, with the speech coding parameters high N position bit possibility value replacement of the high N position bit value usefulness metric minimum of the value of declaring firmly; Afterwards, condition judgment unit 320 judge again speech coding parameters firmly the value of declaring whether satisfy the condition that presets.

Here the high N position bit of each speech coding parameters may value be a 2N value of the high N position bit of speech coding parameters, wherein N is less than or equal to the bit of speech coding parameters, as speech coding parameters 5 bits are arranged, then its high 3 bits possibility value is 8 values between 000 to 111.Metric mainly is meant the transition probability of speech coding parameters and last subframe in the current subframe, and with the transition probability of other speech coding parameters, and the bit error probability of this possibility value, and the bit correct probability equiprobability factor of this probable value with.

The corresponding metric of each high N position bit possibility value, the high N position bit value of the value of declaring may the value replacement with the high N position bit of metric minimum firmly with speech coding parameters for substituting unit 420.

In another specific embodiment, substituting unit 420 can comprise:

A high position is the value of declaring acquiring unit firmly, is used to obtain the high N position of speech coding parameters than the ultrahard value of declaring;

The value of declaring acquiring unit is when obtaining firmly for a high position, and high N position is than the ultrahard value of declaring in the value of declaring firmly by extracting speech coding parameters, and wherein N is less than or equal to the speech coding parameters bit of the value of declaring firmly.

Channel metric acquiring unit, be used for high N position is compared than the value of the ultrahard value of declaring with high N position bit possibility value corresponding bits position, obtain the identical or different comparative result of value of bit, according to comparative result, and first corresponding relation of bit channel information that presets and metric, obtain the channel metric of each bit, and with the channel metric sum of each bit as this high N position bit may the value correspondence the channel metric; The bit channel information comprises the combination of channel decoding soft value and following any or a plurality of information of bit: average error bit rate, speed sign;

Information source metric acquiring unit is used for according to information source information that presets and second corresponding relation of n rank probability, obtains the information source metric of a n rank probability as this high N position bit possibility value correspondence; N is more than or equal to 0; Information source information comprises that high N position bit may value and the combination of following any or a plurality of information: the high N position of the speech coding parameters of last subframe is than the ultrahard value of declaring, the high-order N of other speech coding parameters of last subframe is than the ultrahard value of declaring, the high N position of this speech coding parameters of other speed is than the ultrahard value of declaring, the speech coding parameters of last subframe is transformed into high N position with the speech coding parameters same rate of current subframe than the ultrahard value of declaring, the correlation of spectral line and last subframe spectral line in the current subframe;

Be appreciated that second corresponding relation can comprise (following is that transition probability is the example explanation with 1 rank probability): the speech coding parameters of last subframe is to the transition probability of current this speech coding parameters of subframe; Other speech coding parameters of last subframe are to the transition probability of current this speech coding parameters of subframe; The correlation of current subframe spectral line and last subframe spectral line; Transition probability between this speech coding parameters of other speed and the same speech coding parameters of current subframe; Last subframe transition probability when current subframe is carried out switching rate etc.

The metric acquiring unit, be used for high N position bit may value the channel metric of correspondence and information source metric and as this high N position bit may the value correspondence metric.

Wherein, information source metric acquiring unit can comprise:

The odd even judging unit is used to judge that current subframe is even number subframe or odd number subframe;

Odd number subframe metric obtains the unit, and being used for working as the current subframe of odd even judgment unit judges is the odd number subframe, then according to the corresponding relation of odd number subframe information source information in second corresponding relation and n rank probability, obtains a n rank probability;

Even number subframe metric obtains the unit, and being used for working as the current subframe of odd even judgment unit judges is the even number subframe, then according to the corresponding relation of even number subframe information source information in second corresponding relation and n rank probability, obtains a n rank probability;

Here second corresponding relation comprises: the corresponding relation of the corresponding relation of odd number subframe information source information and n rank probability and even number subframe information source information and n rank probability.

Need to prove the implementation method of each unit and the reciprocal process associated description in can reference method embodiment in Fig. 9 of the present invention and the decoding system shown in Figure 10.

One embodiment of the invention also provides a kind of receiving terminal, comprises the decoding system shown in any as Fig. 9 and Figure 10.

Another embodiment of the present invention also provides a kind of message source and channel decoding system, and its structure can comprise as Figure 11:

The second soft value acquiring unit 50 is used to obtain the channel decoding soft value of received signal frame;

The second soft value acquiring unit 50 can extract the channel decoding soft value of received signal frame from the result of channel decoding; Or, the external information of extraction demodulation bit from the result of channel decoding, the relation according between coding multinomial and the coding output sequence reaches the external information of demodulation bit and the relation between the demodulation bit soft information, obtains the channel decoding soft value of received signal frame.

The second speech frame judging unit 60 is used to judge whether the received signal frame that the second soft value acquiring unit 50 obtains is correct speech frame;

Second amending unit 70 is used for judging that when the second speech frame judging unit 60 the received signal frame is not correct speech frame that then the speech coding parameters at least one received signal frame in the channel decoding soft value carries out corresponding correction;

The speech coding parameters that is appreciated that the received signal frame comprises line spectrum parameter L SF; Second amending unit 70 smoothly calculates the line spectrum parameter L SF in the channel decoding soft value with the state parameter of the last signal frame of current storage, with the correction value of the value after smoothly calculating as line spectrum parameter L SF when the LSF parameter is revised.

Second stores output unit 80, is used for second amending unit, 70 revised speech coding parameters are stored and export as the channel decoding result.

Need to prove the implementation method of each unit and the reciprocal process associated description in can reference method embodiment in the decoding system of Figure 11 of the present invention.

One embodiment of the invention also provides a kind of receiving terminal, comprises decoding system as described in Figure 11.

As seen, the embodiment of the invention provides channel decoding method, system and device, and wherein interpretation method comprises: the channel decoding soft value of obtaining the received signal frame; When determining that at least one speech coding parameters need be revised in the channel decoding soft value, then need the speech coding parameters of correction to carry out corresponding correction, and revised speech coding parameters is stored and exports as the channel decoding result determining.Like this after channel decoding, through after revising, can both meet separately attributive character and also less to each speech coding parameters in the decode results, thereby improve the accuracy of decode results the voice signal influence.

One of ordinary skill in the art will appreciate that all or part of step in the whole bag of tricks of the foregoing description is to instruct relevant hardware to finish by program, this program can be stored in the computer-readable recording medium, and storage medium can comprise: read-only memory (ROM), random-access memory (ram), disk or CD etc.

More than to channel decoding method, system and device that the embodiment of the invention provided, be described in detail, used specific case herein principle of the present invention and execution mode are set forth, the explanation of above embodiment just is used for helping to understand method of the present invention and core concept thereof; Simultaneously, for one of ordinary skill in the art, according to thought of the present invention, the part that all can change in specific embodiments and applications, in sum, this description should not be construed as limitation of the present invention.

Claims

1. a channel decoding method is characterized in that, comprising:

Obtain the channel decoding soft value of received signal frame;

2. the method for claim 1 is characterized in that,

If described speech coding parameters comprises constant codebook excitations FG, the described condition that presets comprises: the high order bit of the constant codebook excitations FG of the described current subframe value of declaring firmly is no more than a FG value that presets, and the FG scope that presetting of the constant codebook excitations FG difference of the last subframe of the constant codebook excitations FG of current subframe and storage; And/or the constant codebook excitations FG of described current subframe firmly the value of declaring be no more than the 2nd FG value that presets;

If described speech coding parameters comprises adaptive codebook excitation AG, the described condition that presets comprises: the adaptive codebook of described current subframe excitation AG firmly the difference of the value of declaring and the adaptive codebook excitation AG of the last subframe that stores less than an AG value that presets, and/or adaptive codebook encourage AG firmly the continuous number of times that does not reduce of the value of declaring less than the threshold value that presets;

If described speech coding parameters comprises adaptive codebook index AI, the described condition that presets comprises: the difference of the inverse quantization value of the value of declaring is in the AI scope that presets firmly for adaptive codebook index AI in described current subframe and last or two subframes, and last signal frame is steady.

3. method as claimed in claim 1 or 2 is characterized in that, if the received signal frame is not correct speech frame, then value of declaring firmly of at least one speech coding parameters also comprises before whether satisfying the condition that presets in judging current subframe:

Obtain the metric of the high N position bit possibility value correspondence of described speech coding parameters, with the high N position bit possibility value replacement of the high order bit value usefulness metric minimum of the value of declaring firmly of described speech coding parameters, wherein N is less than or equal to the speech coding parameters number of bits of the value of declaring firmly.

4. method as claimed in claim 3 is characterized in that, the described metric that obtains the high N position bit possibility value correspondence of speech coding parameters comprises:

Obtain with the corresponding high N of described speech coding parameters position than the ultrahard value of declaring; The value of described high N position than corresponding bits position in ultrahard value of declaring and the high N position bit possibility value compared, obtain the identical or different comparative result of value of bit;

According to described comparative result, and first corresponding relation of bit channel information that presets and metric, obtain the channel metric of each bit, and with the channel metric sum of each bit as this high N position bit may the value correspondence the channel metric, wherein, described bit channel information comprises the combination of channel decoding soft value and following any or a plurality of information of bit: average error bit rate, speed sign;

Second corresponding relation according to information source information that presets and n rank probability, obtain the information source metric of a n rank probability as this high N position bit possibility value correspondence, wherein, described n is more than or equal to 0, described information source information comprises that described high N position bit may value and the combination of following any or a plurality of information: the high N position of the speech coding parameters of last subframe is than the ultrahard value of declaring, the high N position of other speech coding parameters of last subframe is than the ultrahard value of declaring, the high N position of this speech coding parameters of other speed is than the ultrahard value of declaring, the speech coding parameters of last subframe is transformed into high N position with the speech coding parameters same rate of current subframe than the ultrahard value of declaring, the correlation of spectral line and last subframe spectral line in the current subframe;

With described high N position bit may value the channel metric of correspondence and information source metric and as this high N position bit may the value correspondence metric.

5. method as claimed in claim 4 is characterized in that, and is described according to described comparative result, and first corresponding relation of bit channel information that presets and metric, and the channel metric that obtains each bit comprises:

If described comparative result is that the value of bit is inequality,, obtain the channel metric of a bit error probability as this bit then according to the corresponding relation of bit channel information in described first corresponding relation and bit error probability;

If described comparative result is that the value of bit is identical,, obtain the channel metric of a bit correct probability as this bit then according to the corresponding relation of bit channel information in described first corresponding relation and bit correct probability;

Comprise in described first corresponding relation: the corresponding relation of the corresponding relation of bit channel information and bit correct probability and bit channel information and bit error probability.

6. method as claimed in claim 4, it is characterized in that, when described speech coding parameters is adaptive codebook index AI, second corresponding relation of information source information that described basis presets and n rank probability, obtain a n rank probability as this high N position bit may the value correspondence the information source metric specifically comprise:

Judge that described current subframe is even number subframe or odd number subframe, if the odd number subframe then according to the corresponding relation of odd number subframe information source information in described second corresponding relation and n rank probability, obtains a n rank probability; If the even number subframe then according to the corresponding relation of even number subframe information source information in described second corresponding relation and n rank probability, obtains a n rank probability;

Described second corresponding relation comprises: the corresponding relation of the corresponding relation of odd number subframe information source information and n rank probability and even number subframe information source information and n rank probability.

7. method as claimed in claim 1 or 2 is characterized in that,

If the described speech coding parameters that needs to revise is constant codebook excitations FG, described to determining that the speech coding parameters that need revise carries out corresponding correction and comprises: as to reduce constant codebook excitations FG, or the value of the bit of channel decoding soft value absolute value minimum in the high order bit of constant codebook excitations FG is changed to 0;

If it is the described speech coding parameters that needs to revise is adaptive codebook excitation AG, described to determining that the speech coding parameters that need revise carries out corresponding correction and comprises: as to reduce adaptive codebook excitation AG to the value that presets;

If it is the described speech coding parameters that needs to revise is adaptive codebook index AI, described to determining that the speech coding parameters that need revise carries out corresponding correction and comprises: the adaptive codebook index AI that replaces current subframe with the adaptive codebook index AI of last subframe.

8. the method for claim 1 is characterized in that, the described channel decoding soft value of obtaining the received signal frame specifically comprises:

From the result of channel decoding, extract the channel decoding soft value of received signal frame; Or,

Extract the external information of demodulation bit from the result of channel decoding, the relation according between coding multinomial and the coding output sequence reaches the external information of described demodulation bit and the relation between the demodulation bit soft information, obtains the channel decoding soft value of received signal frame.

9. a channel decoding method is characterized in that, comprising:

Obtain the channel decoding soft value of received signal frame;

10. method as claimed in claim 9 is characterized in that, described speech coding parameters comprises line spectrum parameter L SF;

Describedly the speech coding parameters of determining at least one received signal frame in the described channel decoding soft value is carried out corresponding correction comprise: the state parameter with the last signal frame of current storage smoothly calculates the line spectrum parameter L SF in the described channel decoding soft value, with the correction value of the value after the level and smooth calculating as described line spectrum parameter L SF.

11., it is characterized in that the described channel decoding soft value of obtaining the received signal frame specifically comprises as claim 9 or 10 described methods:

12. a channel decoding system is characterized in that, comprising:

13. system as claimed in claim 12 is characterized in that, also comprises:

Substituting unit, be used for when the described received signal frame of the described first speech frame judgment unit judges is not correct speech frame, obtain the metric of the possible value correspondence of high N position bit of described speech coding parameters, with the high N position bit possibility value replacement of the high N position bit value usefulness metric minimum of the value of declaring firmly of described speech coding parameters; Wherein N is less than or equal to the speech coding parameters number of bits of the value of declaring firmly.

14. system as claimed in claim 13 is characterized in that, described substituting unit comprises:

A high position is the value of declaring acquiring unit firmly, be used to obtain with the corresponding high N of described speech coding parameters position than the ultrahard value of declaring;

Channel metric acquiring unit, be used for described high N position is compared than the value of the ultrahard value of declaring with high N position bit possibility value corresponding bits position, obtain the identical or different comparative result of value of bit, according to described comparative result, and first corresponding relation of bit channel information that presets and metric, obtain the channel metric of each bit, and with the channel metric sum of each bit as this high N position bit may the value correspondence the channel metric; Described bit channel information comprises the combination of channel decoding soft value and following any or a plurality of information of bit: average error bit rate, speed sign;

Information source metric acquiring unit is used for according to information source information that presets and second corresponding relation of n rank probability, obtains the information source metric of a n rank probability as this high N position bit possibility value correspondence; Described n is more than or equal to 0; Described information source information comprises that described high N position bit may value and the combination of following any or a plurality of information: the high N position of the speech coding parameters of last subframe is than the ultrahard value of declaring, the high N position of other speech coding parameters of last subframe is than the ultrahard value of declaring, the high N position of this speech coding parameters of other speed is than the ultrahard value of declaring, the speech coding parameters of last subframe is transformed into high N position with the speech coding parameters same rate of current subframe than the ultrahard value of declaring, the correlation of spectral line and last subframe spectral line in the current subframe;

The metric acquiring unit, be used for described high N position bit may value the channel metric of correspondence and information source metric and as this high N position bit may the value correspondence metric.

15. system as claimed in claim 14 is characterized in that, described information source metric acquiring unit comprises:

The odd even judging unit is used to judge that described current subframe is even number subframe or odd number subframe;

Odd number subframe metric obtains the unit, and being used for working as the described current subframe of described odd even judgment unit judges is the odd number subframe, then according to the corresponding relation of odd number subframe information source information in described second corresponding relation and n rank probability, obtains a n rank probability;

Even number subframe metric obtains the unit, and being used for working as the described current subframe of described odd even judgment unit judges is the even number subframe, then according to the corresponding relation of even number subframe information source information in described second corresponding relation and n rank probability, obtains a n rank probability;

16. a channel decoding system is characterized in that, comprising:

17. system as claimed in claim 16, it is characterized in that, described second amending unit specifically is used for: the line spectrum parameter L SF to described channel decoding soft value smoothly calculates with the state parameter of the last signal frame of current storage, with the correction value of the value after smoothly calculating as described line spectrum parameter L SF.

18. a receiving terminal is characterized in that, comprises each described channel decoding system of claim 12 to 15.

19. a receiving terminal is characterized in that, comprises each described channel decoding system of claim 16-17.