CN101409075A - Method for transforming and quantifying line spectrum pair coefficient of G.729 standard - Google Patents
Method for transforming and quantifying line spectrum pair coefficient of G.729 standard Download PDFInfo
- Publication number
- CN101409075A CN101409075A CNA2008101621579A CN200810162157A CN101409075A CN 101409075 A CN101409075 A CN 101409075A CN A2008101621579 A CNA2008101621579 A CN A2008101621579A CN 200810162157 A CN200810162157 A CN 200810162157A CN 101409075 A CN101409075 A CN 101409075A
- Authority
- CN
- China
- Prior art keywords
- point
- code book
- coefficient
- local area
- code word
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The invention relates to a method used for converting and quantifying line spectrum pair coefficients in G729 standard. The traditional method is complex in operation and influential to speech process efficiency. The converting method of the invention comprises the steps as follows: 64 dots in a code book are divided into 10 sections; 10 line spectrum pair coefficients qis are respectively corresponding to the 10 sections; the maximum dots in the 10 sections are compared with the 10 qis; and the obtained 10 dots which are best matched the corresponding 10 qis are calculated to obtain 10 LSF coefficients omegais. The quantifying method comprises the steps as follows: firstly, all the 10-dimensional code words in the original code book are converted into 5-dimensioanl code words; then, a vector consisting of the 10 LSF coefficients is converted into a 5-dimensional target vector; the square error between each code word in the code book and the target vector is calculated; and finally, the quantified result is obtained. In the method, operation complexity is greatly reduced. Meanwhile, the tone of the obtained speech is approximately the same as the tone of the speech obtained by the traditional method; moreover, the tone obtained by the method of the invention is better than the tone obtained by the traditional method.
Description
Technical field
The invention belongs to the encoding and decoding speech field, relate to a kind of LSP of standard (LineSpectrum Pairs, line spectrum pair) coefficient G.729 supported and change fast and to the method for its quantification to LSF (Line Spectral Frequencies, line spectral frequencies) coefficient.
Background technology
G.729 the voice coding standard adopts " CS-ACELP " (the Conjugate-Structured AlgebraicCode Excited Linear Prediction, conjugated structure algebraic codebook excited linear prediction encoding) algorithm.It combines the advantage of waveform coding and parameter coding, based on adaptive predicative coding, has adopted technology such as vector quantization, synthesis analysis and perceptual weighting.This standard application comprises IP phone, radio communication, digital satellite system and digital private circuit in extensive fields.The bit rate of this standard is 8kbps, and the quality of speech signal that its processing obtains is good, and shortcoming is the algorithm complexity, and this handles for the real-time of voice signal and makes troubles.Correlation technique canonical reference ITU.ITU-T Recommendation is speech at 8 kbits/s using conjugate-structure algebraic-code-excitedlinear-prediction (CS-ACELP) [G] .ITU:[s.n. G.729.Codingof], 1996. in present stage, DSP (digital signal processor, digital signal processor) is generally adopted in the realization of encoding and decoding speech.Whole cataloged procedure comprises: the interpolation conversion of the conversion of pre-service, auto-correlation calculating, LP coefficient calculations, LP-LSP coefficient, the conversion of LSP coefficient and quantification, LSP coefficient etc., understand weighting, the search of open loop tone, self-adapting code book search, fixed codebook search, the quantification of gain, the correction of storer etc.Wherein the algorithm complex of LSP coefficient conversion and quantized segment is than higher, and its calculated amount shared weight in the middle of whole encoding-decoding process is bigger.
In the quantification of LSP coefficient is calculated, 10 LSP coefficient q
iAt first be transformed into the expression ω that uses LSF in normalized frequency field [0, π]
iQuantize.These 10 LSF coefficients satisfy order characteristic 0＜ω
1＜ω
2＜...＜ω
10＜π.In standard G.729, its calculation process is exactly in fact to search out 10 points that mate the most with these 10 LSP coefficients in the code book of a cosine value that contains 64 points (between 0 to π).In this code book, the 1st o'clock to 64 orders of pressing from big to small are arranged in order.Its computation process is as follows: at first, the minimum value from code book (i.e. the 64th point) beginning is with q
10Relatively big or small.If more than or equal to q
10, then the 64th point promptly is q
10Pairing point; If less than q
10, then the front a bit continues to compare with it, up to finding just more than or equal to q
10Point till, corresponding point is q
10Mate most; After corresponding point find, just obtained ω through a series of computings again
10After the point search of other 9 fronts such as point finishes, follow more preceding the restarting of point found in code book again and search for.After corresponding point find, just obtained ω through a series of computings again
iAll find 10 ω up to 10 points
iAfter all calculating, convert.
Above-mentioned this transfer process is calculated too loaded down with trivial details, shows:
(1) adopts full method of searching for, make that 64 some basic needs in the code book are all searched for one time;
(2) each ω
iThe ω that the front such as all needs
iAfter obtaining, searching and computing just can enter the code book search.
Many places apply to this transfer algorithm in encoding and decoding standard G.729, and the time of accumulating waste can be more.
10 LSP coefficients are converted to after the LSF coefficient, enter quantization stage.Have two kinds of methods in LSF coefficient quantization field at present, be described below:
A kind ofly adopt the method for full search both according to original computing method of standard G.729, calculate square error and obtain, input vector must compare with each code word.The quantizing process algorithm complex mainly is the search that concentrates on the one-level code book, and the one-level code book has the code word of 128 10 dimensions, calculate the square error of target vector and each code word, relatively obtains the code word of least squares error at last.Here subtraction has used 128 * 10+127 totally 1407 times, and multiplication has used totally 128 * 10 for 1280 times, and addition has used totally 128 * 9 for 1152 times.Full search vector quantization encoding must just can obtain final matching results through behind a large amount of plus-minus multiplications.Though this method can very accurately find the code word that will look for, the search procedure calculated amount is too heavy.
Another kind method is: at first calculating line spectral is to each code word vector in each code book of parameter vector and the distance of true origin, according to sorting of each code word vector in each code book and true origin, make up apart from code book and location index thereof apart from size; In distance code book and location index thereof, search for code word vector with vector distance minimum to be quantified according to minimum distance criterion again, obtain a preliminary Search Results; In the certain limit of this preliminary Search Results, search for again, obtain final Search Results according to minimum distance criterion.This algorithm can reduce the complexity of search procedure in original standard, and raises the efficiency greatly.But there is a shortcoming, promptly when correct code word from the distance of initial point and target vector when the distance of initial point differs bigger, in other words when correct code word is not within the certain limit of preliminary Search Results, then can make a mistake, this correct code word just can not be found like this.After preliminary Search Results found, the expanded range of search again is a lot, and this can solve the generation of above-mentioned mistake, but this has lost the high advantage of original method counting yield.
Therefore, the quantification that is transformed into from the LSP coefficient to the LSF coefficient, by above-mentioned these computing method, it is many to lose time, and this certainly will will have influence on the efficient of speech processes.
Summary of the invention
Purpose of the present invention is exactly at the deficiencies in the prior art, provides LSP coefficient in a kind of G.729 standard to the conversion and to the method for its quantification fast of LSF coefficient.This method greatly reduces the complexity of algorithm, has improved efficient, and the tonequality after optimizing is constant substantially.
In the middle of the conversion Calculation process of LSF coefficient, pairing 10 some position distribution are even in 64 points of 10 LSP coefficients in the middle of 0 to π at the LSP coefficient.The inventive method is utilized this characteristic of LSP coefficient, will enter search simultaneously behind 64 some subregions.
The inventive method comprises conversion method and the quantization method of LSP coefficient to the LSF coefficient.
The LSP coefficient is as follows to the transfer process of LSF coefficient:
(1) at first be that code book is carried out initialization, from big to small tactic 64 points are divided into 10 districts, wherein two districts of head and the tail respectively are 8 points, and middle 8 districts respectively are 6 points; 10 line spectrum pair coefficient q
i(i=1 ..., 10) and corresponding with 10 districts respectively.
(2) the maximum of points X in 10 districts of code book
MAX(being the 1st point in each district) be while and 10 line spectrum pair coefficient q respectively
i(i=1 ..., 10) compare:
A. if X
MAXQ greater than local area
i, X in the local area then
MAXBack point continue q with local area
iCompare, up to the q that in local area, finds a point less than local area
iTill; The previous point of the point that then finds is q
iThe point that mates most;
B. if X
MAXQ less than local area
i, then get the minimum point X in previous district
MINThe q of (being last point in the previous district, is the 6th point or the 8th point) and local area
iCompare, if also less than q
i, then continue to use X
MINPreceding 1 point and q
iCompare, up to finding a point more than or equal to q
i, this point is q
iThe point that mates most;
C. if X
MAXEqual the q of local area
i, then this point is q
iThe point that mates most.
(3) will find 10 with 10 corresponding q
iThe point that mates most is by calculating 10 LSF coefficient ω
iComputing method are according to the method in the existing G.729 standard.
The present invention utilizes pairing 10 uniform characteristics of some position distribution in 64 points of 10 LSP coefficients in the middle of 0 to π, proposed 64 points are divided into 10 districts, and the new method of simultaneously code word of 10 LSP coefficients in subregion being searched for entirely.New method has solved original algorithm and has needed the full shortcoming of searching for of whole code book, and in new method, general code word of searching for half in the correspondence district can find 10 whole points; The LSP coefficient that new method has solved front in original algorithm need wait the LSP coefficient of back to search out the shortcoming that the result just can enter search afterwards, and 10 LSP coefficients all enter search simultaneously in the new method; From these 2, the present invention compares with original algorithm, has obviously reduced algorithm complex, has improved the conversion efficiency of LSP coefficient to the LSF coefficient.G.729 in the standard, similarly transfer algorithm does not occupy the minority really for it, only with the LSP coefficient to this process of conversion of LSF coefficient as an example, this quick idea of transformation is described here.
In the middle of the process of LSF coefficient quantization, analyze the shortcoming of existing two kinds of methods, find that correct code word is smaller from the target vector distance, can not represent that correct code word differs smaller from the distance and the target vector of initial point from the distance of initial point.Code word in the code book and target vector may differ greatly by coordinate figure on the coordinate of indivedual dimensions, and both are equal or approaching from the initial point distance, and such code word obviously is not the code word that will look for.The code word of this mistake, if the coordinate that the coordinate on certain one dimension is corresponding with target vector is far short of what is expected, then generally in another dimension or the coordinate of the coordinate of multidimensional and target vector correspondence also can be far short of what is expected in addition, but it from the distance of initial point may be than correct code word from the initial point distance more near the distance of target vector from initial point.Therefore because the existence of these points may cause correct code word not within the scope of search again, and search less than.
The concrete grammar that quantizes in the inventive method is:
(4) at first existing one-level code book in the standard is carried out the initialization conversion, with the code word (x of all 10 dimensions in original code book
0, x
1, x
2, x
3, x
4, x
5, x
6, x
7, x
8, x
9) convert to 5 the dimension code word (y
0, y
1, y
2, y
3, y
4), y
i=x
2i 2+ x
2i+1 2, y
iFor this code word at x
2iAnd x
2i+1The distance of place plane projection square.
(5) vector (ω that 10 LSF coefficients are constituted
0, ω
1, ω
2, ω
3, ω
4, ω
5, ω
6, ω
7, ω
8, ω
9) be converted into the 5 target vector (z that tie up according to the method in the step (4)
0, z
1, z
2, z
3, z
4), z
i=ω
2i 2+ ω
2i+1 2
(6) each code word in the calculating code book and the square error between the target vector specifically are square y of 5 projector distances in the code book after the conversion
iSquare z with 5 of target vector corresponding projector distances
iSubtract each other back square, again 5 square value additions are obtained square error; The code word of least squares error correspondence is optimal codes.
(7) according to Search Results to the one-level code book, finish remaining calculating, get quantized result to the end.Computing method are according to the method in the existing G.729 standard.
The present invention utilize code word to 5 a continuous plane, bidimensional place projection distance square with original 10 the dimension code words be converted into 5 the dimension code words.This is unlike in the method two, and 10 dimension data are converted into 1 distance value.Part code word and target vector may differ greatly by coordinate figure on the coordinates of indivedual dimensions in the code book, but it from the distance of initial point very near the distance of target vector from initial point, here error of existing of each dimension be 1 distance value can not embody.And in the present invention, the every error between these wrong code words and the target vector then is embodied in the middle of 5 projector distances.Then the present invention can avoid occurring choosing the generation of the situation of these wrong code words.Dimensionality reduction makes target vector reduce greatly the calculated amount of the full search procedure of whole code book.
Tonequality after the optimization of the present invention is the same substantially with the tonequality that obtains with method one, and better than the tonequality that obtains with method two.
Embodiment
The method of conversion of line spectrum pair coefficient and quantification in a kind of G.729 standard, wherein conversion method is:
(1) at first be that code book is carried out initialization.Tactic 64 points from big to small are divided into 10 districts, and wherein two districts of head and the tail respectively are 8 points, and middle 8 districts respectively are 6 points; 10 line spectrum pair coefficient q
i(i=1 ..., 10) and corresponding with 10 districts respectively.
The subregion has here made full use of two characteristics that 1st o'clock to 64 order of pressing from big to small of even and this code book of pairing 10 some position distribution in 64 points of 10 LSP coefficients in the middle of 0 to π is arranged in order.
(2) the maximum of points X in 10 districts of code book
MAX(being the 1st point in each district) be while and 10 line spectrum pair coefficient q respectively
i(i=1 ..., 10) compare:
A. if X
MAXQ greater than local area
i, X in the local area then
MAXBack point continue q with local area
iCompare, up to the q that in local area, finds a point less than local area
iTill; The previous point of the point that then finds is q
iThe point that mates most;
B. if X
MAXQ less than local area
i, then get the minimum point X in previous district
MINThe q of (being last point in the previous district, is the 6th point or the 8th point) and local area
iCompare, if also less than q
i, then continue to use X
MINPreceding 1 point and q
iCompare, up to finding a point more than or equal to q
i, this point is q
iThe point that mates most;
C. if X
MAXEqual the q of local area
i, then this point is q
iThe point that mates most.
According to the method described above, general in the correspondence district half code word of search can all find 10 whole points, having solved needs shortcoming that whole code book is searched for entirely in original algorithm; 10 LSP coefficients all enter search simultaneously in the present invention, and the LSP coefficient that has solved front in original algorithm need wait the LSP coefficient of back to search out the shortcoming that the result just can enter search afterwards.
(3) will find 10 with 10 corresponding q
iThe point that mates most calculates 10 LSF coefficient ω
iComputing method are according to the method in the existing G.729 standard.
This method is utilized pairing 10 uniform characteristics of some position distribution in 64 points of 10 LSP coefficients in the middle of 0 to π, proposed 64 points are divided into 10 districts, and the new method of simultaneously code word of 10 LSP coefficients in subregion being searched for entirely.The present invention compares with original algorithm, has obviously reduced algorithm complex, has improved the conversion efficiency of LSP coefficient to the LSF coefficient.
The method that quantizes is:
(4) at first existing one-level code book in the standard is carried out the initialization conversion, with the code word (x of all 10 dimensions in original code book
0, x
1, x
2, x
3, x
4, x
5, x
6, x
7, x
8, x
9) convert to 5 the dimension code word (y
0, y
1, y
2, y
3, y
4), y
i=x
2i 2+ x
2i+1 2, y
iFor this code word at x
2iAnd x
2i+1The distance of place plane projection square.
(5) vector (ω that 10 LSF coefficients are constituted
0, ω
1, ω
2, ω
3, ω
4, ω
5, ω
6, ω
7, ω
8, ω
9) be converted into the 5 target vector (z that tie up according to the method in the step (4)
0, z
1, z
2, z
3, z
4), z
i=ω
2i 2+ ω
2i+1 2
Utilize code word to 5 a continuous plane, bidimensional place projection distance square with original 10 the dimension code words be converted into 5 the dimension code words.Like this can be so that 5 projector distances after the conversion include the most information of 10 coordinate figures of original 10 n dimensional vector ns.This is converted into 1 simple distance value unlike method two with original vector, and this can cause the most information of 10 coordinate figures of original 10 n dimensional vector ns to be lost.
(6) each code word in the calculating code book and the square error between the target vector specifically are square y of 5 projector distances in the code book after the conversion
iSquare z with 5 of target vector corresponding projector distances
iSubtract each other back square, again 5 square value additions are obtained square error; The code word of least squares error correspondence is optimal codes.
Dimensionality reduction make target vector to the full search procedure of whole code book unlike the intensive in the method one.Here the calculating in the initialization procedure of code book was just finished before quantizing, and did not comprise this part calculating in the middle of cataloged procedure.And mainly be the dimensionality reduction process and the quantizing process of target vector here.In whole search procedure, subtraction has used 128 * 5+127 totally 767 times, and multiplication has used 10+128 * 5 totally 650 times, and addition has used 5+128 * 4 totally 517 times.Such calculated amount is half of method one basically, and promptly efficient is doubled.
(7) according to Search Results to the one-level code book, finish remaining calculating, get quantized result to the end.Computing method are according to the method in the existing G.729 standard.
The present invention utilize code word to 5 a continuous plane, bidimensional place projection distance square with original 10 the dimension code words be converted into 5 the dimension code words.5 projector distances that obtain by dimensionality reduction include the most information of 10 coordinate figures of original 10 n dimensional vector ns, and dimensionality reduction makes target vector reduce half to the calculated amount of the full search procedure of whole code book simultaneously.
The present invention has reduced many with respect to conventional method at algorithm complex. The voice that obtain simultaneously Tonequality is substantially the same with the tonequality that obtains with conventional method, and better than the tonequality that obtains with conventional method.
Claims (1)
1.G.729 the method for conversion of line spectrum pair coefficient and quantification comprises conversion method and the quantization method of line spectrum pair coefficient to the line spectral frequencies coefficient in the standard, it is characterized in that:
The line spectrum pair coefficient to the conversion method of line spectral frequencies coefficient is:
(1) at first be that code book is carried out initialization, from big to small tactic 64 points are divided into 10 districts, wherein two districts of head and the tail respectively are 8 points, and middle 8 districts respectively are 6 points; 10 line spectrum pair coefficient q
i(i=1 ..., 10) and corresponding with 10 districts respectively;
(2) the maximum of points X in 10 districts of code book
MAXDifference while and 10 line spectrum pair coefficient q
i(i=1 ..., 10) compare:
A. if X
MAXQ greater than local area
i, X in the local area then
MAXBack point continue q with local area
iCompare, up to the q that in local area, finds a point less than local area
iTill; The previous point of the point that then finds is q
iThe point that mates most;
B. if X
MAXQ less than local area
i, then get the minimum point X in previous district
MINQ with local area
iCompare, if also less than q
i, then continue to use X
MINPreceding 1 point and q
iCompare, up to finding a point more than or equal to q
i, this point is q
iThe point that mates most;
C. if X
MAXEqual the q of local area
i, then this point is q
iThe point that mates most;
(3) will find 10 with 10 corresponding q
iThe point that mates most is by calculating 10 line spectral frequencies coefficient ω
i
The concrete grammar that quantizes is:
(4) at first existing one-level code book in the standard is carried out the initialization conversion, with the code word (x of all 10 dimensions in original code book
0, x
1, x
2, x
3, x
4, x
5, x
6, x
7, x
8, x
9) convert to 5 the dimension code word (y
0, y
1, y
2, y
3, y
4), y
i=x
2i 2+ x
2i+1 2, y
iFor this code word at x
2iAnd x
2i+1The distance of place plane projection square;
(5) vector (ω that 10 line spectral frequencies coefficients are constituted
0, ω
1, ω
2, ω
3, ω
4, ω
5, ω
6, ω
7, ω
8, ω
9) convert to 5 the dimension target vector (z
0, z
1, z
2, z
3, z
4), z
i=ω
2i 2+ ω
2i+1 2
(6) each code word in the calculating code book and the square error between the target vector specifically are square y of 5 projector distances in the code book after the conversion
iSquare z with 5 of target vector corresponding projector distances
iSubtract each other back square, again 5 square value additions are obtained square error; The code word of least squares error correspondence is optimal codes;
(7) according to Search Results, by calculating last quantized result to the one-level code book.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101621579A CN101409075B (en) | 2008-11-27 | 2008-11-27 | Method for transforming and quantifying line spectrum pair coefficient of G.729 standard |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2008101621579A CN101409075B (en) | 2008-11-27 | 2008-11-27 | Method for transforming and quantifying line spectrum pair coefficient of G.729 standard |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101409075A true CN101409075A (en) | 2009-04-15 |
| CN101409075B CN101409075B (en) | 2011-05-11 |
Family
ID=40572055
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2008101621579A Active CN101409075B (en) | 2008-11-27 | 2008-11-27 | Method for transforming and quantifying line spectrum pair coefficient of G.729 standard |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101409075B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104143337A (en) * | 2014-01-08 | 2014-11-12 | 腾讯科技（深圳）有限公司 | Method and device for improving tone quality of sound signal |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| AU2002218501A1 (en) * | 2000-11-30 | 2002-06-11 | Matsushita Electric Industrial Co., Ltd. | Vector quantizing device for lpc parameters |
| CN1244903C (en) * | 2003-10-30 | 2006-03-08 | 北京首信股份有限公司 | Quick algorithm for searching weighted quantized vector of line spectrum in use for encoding voice |
-
2008
- 2008-11-27 CN CN2008101621579A patent/CN101409075B/en active Active
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104143337A (en) * | 2014-01-08 | 2014-11-12 | 腾讯科技（深圳）有限公司 | Method and device for improving tone quality of sound signal |
| CN104143337B (en) * | 2014-01-08 | 2015-12-09 | 腾讯科技（深圳）有限公司 | A kind of method and apparatus improving sound signal tonequality |
| US9646633B2 (en) | 2014-01-08 | 2017-05-09 | Tencent Technology (Shenzhen) Company Limited | Method and device for processing audio signals |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101409075B (en) | 2011-05-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7805292B2 (en) | Method and apparatus for audio transcoding | |
| JP6316398B2 (en) | Apparatus and method for quantizing adaptive and fixed contribution gains of excitation signals in a CELP codec | |
| KR100788706B1 (en) | Method for encoding and decoding of broadband voice signal | |
| CN101057275B (en) | Vector conversion device and vector conversion method | |
| CN1890714B (en) | Optimized multiple coding method | |
| KR20000010994A (en) | Audio signal coding and decoding methods and audio signal coder and decoder | |
| US11942102B2 (en) | Pyramid vector quantizer shape search | |
| US20100174539A1 (en) | Method and apparatus for vector quantization codebook search | |
| KR20120043160A (en) | Multi-stage quantizing method and device | |
| WO2007070007A1 (en) | A method and system for extracting audio features from an encoded bitstream for audio classification | |
| WO2010091554A1 (en) | Method and device for pitch period detection | |
| US20060184362A1 (en) | Speech analyzing system with adaptive noise codebook | |
| AU2014350366A1 (en) | Encoder for encoding an audio signal, audio transmission system and method for determining correction values | |
| WO2007046060A2 (en) | Compression and decompression of data vectors | |
| CN103069483B (en) | Encoder apparatus and encoding method | |
| US8751225B2 (en) | Apparatus and method for coding signal in a communication system | |
| CN100541609C (en) | A kind of method and apparatus of realizing open-loop pitch search | |
| US8335260B2 (en) | Method and device for vector quantization | |
| CN101937680A (en) | Vector quantization method for sorting and rearranging code book and vector quantizer thereof | |
| CN101409075B (en) | Method for transforming and quantifying line spectrum pair coefficient of G.729 standard | |
| EP1771841B1 (en) | Method for generating and using a vector codebook, method and device for compressing data, and distributed speech recognition system | |
| WO2024021747A1 (en) | Sound coding method, sound decoding method, and related apparatuses and system | |
| EP0868031B1 (en) | Signal coding method and apparatus | |
| CN110050304A (en) | Method, the encoder and decoder of coefficient are indicated for handling envelope | |
| CN105741853A (en) | Digital speech perception hash method based on formant frequency |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant |