CN103137135B - LPC coefficient quantization method and device and multi-coding-core audio coding method and device - Google Patents
LPC coefficient quantization method and device and multi-coding-core audio coding method and device Download PDFInfo
- Publication number
- CN103137135B CN103137135B CN201310027233.6A CN201310027233A CN103137135B CN 103137135 B CN103137135 B CN 103137135B CN 201310027233 A CN201310027233 A CN 201310027233A CN 103137135 B CN103137135 B CN 103137135B
- Authority
- CN
- China
- Prior art keywords
- coding
- vector quantization
- audio
- signal
- sound signal
- 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.)
- Active
Links
Landscapes
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
The invention relates to a method and a device for quantizing LPC coefficients for audio signal coding. The method comprises the following steps: s1, determining the type of the input audio signal based on a predetermined rule; s2, linear prediction processing is carried out on the input audio signal, and an LPC coefficient is calculated; and S3, for different audio signal types, applying a vector quantization code book matched with the audio signal type to carry out vector quantization on the LPC coefficients. The invention also relates to a multi-coding-core audio coding method and a device adopting the LPC coefficient quantization method and device. The invention quantizes LPC coefficients based on audio signal classification, is applied to a multi-coding-core coding algorithm which has at least one coding core and uses linear prediction LPC to code at least two types of audio signals, and can further improve the quantization precision of prediction parameters of an internal linear prediction coding module, thereby improving the efficiency of the whole digital audio coding algorithm and the subjective sound quality of a coder.
Description
Technical field
The present invention relates to Digital Audio Coding Technology, more particularly, relate to a kind of LPC coefficient quantization method and apparatus for audio-frequency signal coding and a kind of odd encoder core audio coding method and equipment.
Background technology
In digital audio encoding, because sound signal is very complicated, generally comprise music class signal, voice class signal and mixing class signal etc., some audio coding algorithms such as MPEG-1, MPEG-2, MPEG-4, Dolby AC-3 and DTS etc. operates mainly in high code check high-quality, when the code efficiency for voice class signal under low bit-rate is lower; And other ITU G series standard encryption algorithm is mainly for low bit-rate voice signal, for broadband signal then code efficiency decline.In order to unanimously obtain higher code efficiency to all types of sound signal; general needs adopts the hybrid coding structure with Multi-encoding kernel, if the AMR-WB+(of 3GPP is see 3GPP TS26.290: " Audio codec processing functions; Extended AMR Wideband codec; Transcoding functions ") and MPEG-D USAC(is see ISO/IEC DIS23003-3-Information technology--MPEG audiotechnologies--Part3:Unified speech and audio coding ") etc.In these hybrid coding algorithms, there is different compression algorithm process to each sound signal type, expect that integrated encode performance is improved.
In AMR-WB+, ACELP(Algebraic Code ExcitedLinear Prediction is adopted for voice signal, algebraic code-excited linear is predicted) coding core, TCX(Transform Coded Excitation is generally adopted for mixing class and music class signal, change code excited) core of encoding, two kinds of coding cores all apply LPC(Linear Predictive Coding, linear predictive coding)) technology describes the short-term spectral envelope of voice, is thus a critical problem in voice coding to the high effective quantization of LPC coefficient.Because the dynamic range of LPC coefficient is larger, for the consideration of composite filter stability and quantitative efficiency, LPC coefficient is converted into re-quantization after the parameter of other form of mathematically equivalent usually, common representation is ISF(Immittance Spectral Frequency, immittance spectral frequencies coefficient) or LSF(LineSpectral Frequency, line spectral frequency parameters).LSF is as a kind of frequency domain parameter of LPC coefficient, better quantize and interpolation characteristic because it has, LPC coefficients conversion is often LSF parameter by voice coding end, and then LSF parameter is carried out quantizing (generally adopting vector quantization technology), tone decoding end carries out re-quantization and obtains the LSF parameter after quantizing, and LSF parameter is converted to LPC coefficient again, therefore LSF is widely used in based on LPC voice coding.
At MPEG-D USAC(Unified Speech and Audio Coding, unified voice/audio coding) in coding, for music class signal, adopt efficient AAC(AdvancedAudio coding, Advanced Audio Coding) coding; For voice signal, general employing ACELP class coding core; For mixing class signal, general employing TCX class coding core.As AMR-WB+, in MPEG-D USAC hybrid coding structure, ACELP and TCX coding core can share LPC coding techniques.
Although the odd encoder core audio coding algorithms such as AMR-WB+ and MPEG-D USAC start have the type to input audio signal to analyze, for dissimilar, adopt different coding core, obtain comprehensive optimum coding efficiency.In AMR-WB+, voice signal class adopts ACELP coding core, and music class and mixing class signal adopt TCX coding core; In MPEG-D USAC, voice class signal adopts ACELP coding core, and mixing class adopts TCX coding core, and music class adopts AAC coding core.All ACELP and TCX is have employed in these two kinds of odd encoder core audio coding algorithms of AMR-WB+ and MPEG-D USAC, and these two coding cores can share a linear prediction LPC technology, and be all generally be after LSF composes parameter to LPC coefficients conversion, carry out vector quantization coding again, and adoptable vector quantization method has a variety of, such as, the applying date is on July 17th, 2012, application number is 201210246780.9, name is called that the Chinese patent application of " for voice signal LPC coefficient being carried out to the method and system of multi-stage vector quantization " just discloses a kind of multilevel vector quantization method, but the code book that these vector quantization methods generate does not rely on the type of the digital audio and video signals of input, namely to all sound signals, all only generate a set of vector quantization code book, thus the quantified precision of LPC coefficient is not still very desirable, thus affect the code efficiency of overall digital audio coding algorithms and the subjective sound quality of scrambler.
Summary of the invention
First technical matters that the present invention will solve is, for the above-mentioned defect of prior art, provides a kind of LPC coefficient quantization method and apparatus for audio-frequency signal coding that can improve quantified precision further.
Second technical matters that the present invention will solve is, for the above-mentioned defect of prior art, provide a kind of can improve intra-prediction parameter quantified precision so that improve the efficiency of overall digital audio coding algorithms and the odd encoder core audio coding method of subjective sound quality and encoding device.
The present invention solves the technical scheme that its first technical matters adopt: propose a kind of LPC coefficient quantization method for audio-frequency signal coding, comprise the steps:
S1, determine the type of input audio signal based on predetermined rule;
S2, linear prediction process is performed to input audio signal, calculate LPC coefficient;
S3, for different sound signal types, application and the vector quantization code book of this sound signal type matching carry out vector quantization to described LPC coefficient.
The present invention is above-mentioned in the LPC coefficient quantization method of audio-frequency signal coding,
Described step S2 comprises further:
It is the LSF parameter of equivalence by described LPC coefficients conversion;
Described step S3 comprises further:
For different sound signal types, application carries out vector quantization with the vector quantization code book of this sound signal type matching to described LSF parameter.
The present invention is above-mentioned in the LPC coefficient quantization method of audio-frequency signal coding, and in described step S3, vector quantization adopts multilevel vector quantization method.
The present invention is above-mentioned in the LPC coefficient quantization method of audio-frequency signal coding, and described method also comprised before step S1:
Vector quantization code book needed for the coding build the signal model for different audio signals type is stored in this locality.
The above-mentioned LPC coefficient quantization method for audio-frequency signal coding of the present invention comprises further:
S4, send the coding parameter of vector quantization to multiplexer and be multiplexed in total audio coding frame.
The present invention solves its first technical matters also to propose a kind of LPC coefficient quantization device for audio-frequency signal coding, comprising:
Audio types determination module, for determining the type of input audio signal based on predetermined rule;
Linear prediction processing module, for performing linear prediction process to input audio signal, calculates LPC coefficient;
Spectrum parameter quantification module, for for different sound signal types, applies and carries out vector quantization with the vector quantization code book of this sound signal type matching to described LPC coefficient.
The present invention is above-mentioned in the LPC coefficient quantization device of audio-frequency signal coding, and described linear prediction processing module comprises further:
LPC coefficients calculation block, performs linear prediction process for input audio signal, calculates LPC coefficient;
Equivalency transform module, for by described LPC coefficients conversion being the LSF parameter of equivalence.
The present invention is above-mentioned in the LPC coefficient quantization device of audio-frequency signal coding, described spectrum parameter quantification module is further used for for different sound signal types, and application carries out vector quantization with the vector quantization code book of this sound signal type matching to described LSF parameter.
The above-mentioned LPC coefficient quantization device for audio-frequency signal coding of the present invention also comprises:
Memory module, the vector quantization code book needed for the coding that the signal model for storing for different audio signals type builds.
The above-mentioned LPC coefficient quantization device for audio-frequency signal coding of the present invention also sends the coding parameter of vector quantization to multiplexer and is multiplexed in total audio coding frame.
The present invention solves the technical scheme that its second technical matters adopt: propose a kind of odd encoder core audio coding method, comprise the steps:
A, the type of input audio signal to be analyzed;
B, for multiple sound signal type, adopt corresponding multiple coding core to encode, wherein, sound signal that at least one coding checks at least two sound signal types performs linear predictive coding;
Wherein, described linear predictive coding is quantized LPC coefficient by the above-mentioned LPC coefficient quantization method for audio-frequency signal coding.
The present invention also proposes a kind of odd encoder core audio coding apparatus for solving its second technical matters, comprising:
Audio signal classification processing module, for analyzing the type of input audio signal;
Multiple coding core, for based on sound signal type to corresponding coding audio signal, wherein, sound signal that at least one coding checks at least two sound signal types carries out linear predictive coding;
Wherein, described linear predictive coding is quantized LPC coefficient by the above-mentioned LPC coefficient quantization device for audio-frequency signal coding.
The LPC coefficient quantization method and apparatus of audio-frequency signal coding is used for by the present invention, when LPC parameter quantification in encryption algorithm is encoded, for different audio signals type provides the vector quantization code book mated most separately respectively, when not needing extra audio frequency signal type indication bit expense, parameter improvement quantified precision can be composed to LPC further.And then the odd encoder core audio coding method of this LPC coefficient quantization method and apparatus of employing of the present invention and equipment can improve the code efficiency of binary encoding algorithm, or reduce coding bit rate under same quality, improve the subjective sound quality of scrambler.
Accompanying drawing explanation
Below in conjunction with drawings and Examples, the invention will be further described, in accompanying drawing:
Fig. 1 is the cataloged procedure schematic block diagram of MPEG-D USAC encryption algorithm;
Fig. 2 is the process flow diagram of the LPC coefficient quantization method for audio-frequency signal coding of one embodiment of the invention;
Fig. 3 is the process flow diagram of the LPC coefficient quantization method for audio-frequency signal coding of another embodiment of the present invention;
Fig. 4 is the logic diagram of the LPC coefficient quantization device for audio-frequency signal coding of one embodiment of the invention.
Embodiment
In order to make object of the present invention, technical scheme and advantage clearly understand, below in conjunction with drawings and Examples, the present invention is further elaborated.Should be appreciated that specific embodiment described herein only in order to explain the present invention, be not intended to limit the present invention.
The signal type of sound signal can be divided into 2 classes, 3 classes or more polymorphic type.Can be voice signal, non-speech audio when being divided into 2 class; Being divided into during 3 class can be voice signal, music signal, voice and music mix class signal.LPC coefficient quantization method and apparatus for audio-frequency signal coding of the present invention just adopts different match vector quantization code books to carry out vector quantization to LPC spectral coefficient based on audio signal classification respectively, thus the quantified precision of spectrum parameter can be improved further, the bit number of spectrum required for parameter coding can be reduced in other words under equal accuracy.
Such as, audio signal classification becomes voice class signal, music class signal and voice music to mix class signal by AMR-WB+ odd encoder core hybrid coding algorithm.When the present invention is applied to AMR-WB+, for voice class signal, use the one group vector quantization code book relevant to this signal model to carry out vector quantization, and then carry out ACELP coding; And for music class and voice music mixing class signal, use two groups of different vector quantization code books to quantize respectively separately, then complete TCX process.
Again such as, audio signal classification also becomes voice class signal, music class signal and voice music to mix class signal by MPEG-D USAC odd encoder core hybrid coding algorithm.When the present invention is applied to MPEG-D USAC, because only voice class and music voice mix the signal demand LPC process of class two type, therefore provide the vector quantization code book of coupling to quantize respectively for this two classes signal, and then correspondence carry out ACELP coding or TCX coding.Below will introduce the present invention in detail for MPEG-D USAC encryption algorithm.
Fig. 1 shows the cataloged procedure of MPEG-D USAC encryption algorithm.As shown in Figure 1, MPEG-DUSAC encryption algorithm mainly comprises three phases.Pretreatment stage: in step 110 to input PCM(Pulse Code Modulation, pulse code modulation (PCM)) sound signal carries out resampling, its objective is when input sampling rate is different with coded sample rate, adjustment input sampling rate is to the optimum sampling rate being applicable to coded treatment; In step 130 signal type analysis is carried out, to carry out different coding process for signal with different type to input pcm audio signal; Carry out the SBR(Spectral band replication around mpeg encoded (MPEG Surround) and enhancing to the sound signal through resampling in the step 120, frequency range copies) process.Dissimilar based on sound signal, is admitted to two codings core, i.e. the first code branch (Frequency Domain Coding core) 140 and the second code branch (time domain coding core) 150 through pretreated sound signal.Introduce as front, in MPEG-D USAC, voice class signal adopts ACELP to encode core, and mixing class signal adopts TCX to encode core, and music class signal adopts AAC to encode core.Also namely, enter the first code branch 140 through pretreated music class signal, estimate through tone, block switches control, psychoacoustic model control, filtering, TNS(Temporal Noise Shaping, time domain noise shaped), the process such as M/S coding; Enter the second code branch 150 through pretreated voice class signal, carry out carrying out ACELP coding again after LPC composes parameter quantification process 151; Enter the second code branch 150 through pretreated mixing class signal, carry out carrying out TCX coding again after LPC composes parameter quantification process 151.Liang Ge branch signal out, after aftertreatment 160, through multiplexer by multiplexing for all coding parameters 170, exports total audio coding frame.
Technical scheme of the present invention is mainly reflected in the improvement of LPC being composed to parameter quantification process 151, proposes a kind of LPC coefficient quantization method and apparatus based on audio signal classification, will provide explanation in detail below.About other functional module and the step of the MPEG-D USAC encryption algorithm shown in Fig. 1, the prior art be well known to those skilled in the art, therefore do not repeat them here.
Fig. 2 is according to an embodiment of the invention for the process flow diagram of the LPC coefficient quantization method 200 of audio-frequency signal coding.As shown in Figure 2, the method 200 comprises the steps:
In step 210, determine the type of input audio signal based on predetermined rule.As previously mentioned, the odd encoder core audio coding algorithms such as MPEG-D USAC can adopt different coding core based on different signal types, first signal type analysis will inevitably be carried out to the pcm audio signal of input, and by classification type parameter coding to (shown in label in Fig. 1 130) in compressed bit stream.Such as, in MPEG-USAC, voice class signal, music class signal, music voice is divided into mix class signal three types.Thus, LPC coefficient quantization method of the present invention does not need additionally to increase signal type processing module and signal type indication information again, can parse the type of input audio signal based on predetermined rule from compressed bit stream.Therefore, when LPC coefficient quantization method of the present invention is applied to odd encoder core audio coding algorithms, do not need to increase any overhead in coded frame.
In later step 220, linear prediction process (LPC) is performed to input audio signal, calculate LPC coefficient.To inputting PCM signal, the present invention calculates LPC coefficient by equitable subsection (256 sampling points calculate once as ACELP coding, and TCX coding possibility 256,512 or 1024 sampling points calculate once).
In later step 230, for different sound signal types, application carries out vector quantization with the vector quantization code book of this sound signal type matching to LPC coefficient.The design of the vector quantization code book needed for the present invention can encode for the signal model of different audio signals type in advance respectively, constructs the vector quantization code book mated most with each sound signal type, and is stored in this locality.Due to the vector quantization code book of the corresponding coupling of different audio signals type, therefore application the present invention carries out audio-frequency signal coding needs to have at coding side and decoding end to store these vector quantization code books, can increase certain memory space requirements.
In step 230, vector quantization is carried out to LPC coefficient, known in the art and feasible various vector quantization methods can be adopted, such as, the applying date is on July 17th, 2012, application number is 201210246780.9, multilevel vector quantization method disclosed in the Chinese patent application that name is called " for voice signal LPC coefficient being carried out to the method and system of multi-stage vector quantization ".
Fig. 3 is the process flow diagram of the LPC coefficient quantization method 300 for audio-frequency signal coding according to another specific embodiment of the present invention.As shown in Figure 3, the method 300 comprises the steps:
In step 310, determine the type of input audio signal based on predetermined rule.
In later step 320, linear prediction process is performed to input audio signal, calculate LPC coefficient.
In later step 330, be the LSF parameter of equivalence by LPC coefficients conversion.
In later step 340, for different sound signal types, what application this locality stored carries out vector quantization with the vector quantization code book of this sound signal type matching to LSF parameter.As previously mentioned, the present invention can carry out the design of vector quantization code book in advance respectively for the LSF parameter of different audio signals type, constructs the vector quantization code book mated most that coding needs and is stored in this locality.
In later step 350, the coding parameter of vector quantization is sent to multiplexer and be multiplexed in total audio coding frame, to send receiving end (demoder) to.
The LPC coefficient quantization method of audio-frequency signal coding is used for by the present invention, when LPC parameter quantification in encryption algorithm is encoded, for different audio signals type provides the vector quantization code book mated most separately respectively, when not needing extra audio frequency signal type indication bit expense, parameter improvement quantified precision can be composed further to LPC, thus improve the code efficiency of binary encoding algorithm, or reduce coding bit rate under same quality.
Fig. 4 is according to an embodiment of the invention for the logic diagram of the LPC coefficient quantization device 400 of audio-frequency signal coding.LPC coefficient quantization device 400 for audio-frequency signal coding comprises audio types determination module 410, linear prediction processing module 420, spectrum parameter quantification module 430 and memory module 440.Wherein, audio types determination module 410 is for determining the type of input audio signal based on predetermined rule.As previously mentioned, the odd encoder core audio coding algorithms such as MPEG-D USAC can adopt different coding core based on different signal types, audio signal classification processing module must be provided with and first signal type analysis be carried out to the pcm audio signal of input, and by classification type parameter coding to (shown in label in Fig. 1 130) in compressed bit stream.Thus, audio types determination module 410 can parse the type of input audio signal from compressed bit stream based on predetermined rule, do not need additionally to increase the process of sound signal type analysis and indication information, therefore do not need to increase any overhead in coded frame.Linear prediction processing module 420, for performing linear prediction process to input audio signal, calculates LPC coefficient.Memory module 440 is for storing in this locality for the vector quantization code book needed for the coding constructed by the signal model of different audio signals type.Spectrum parameter quantification module 430 for for different sound signal types, application is local store with the vector quantization code book of this sound signal type matching, vector quantization is carried out to the LPC coefficient that linear prediction processing module 420 calculates.
In specific embodiment, as shown in Figure 4, linear prediction processing module 420 comprises LPC coefficients calculation block 421 and equivalency transform module 422 further.LCP coefficients calculation block 421 pairs of input audio signals calculate LPC coefficient by equitable subsection.LCP coefficients conversion is become the LSF parameter of equivalence by equivalency transform module 422.Further, general parameter quantification module 430 is for different sound signal types, and what application this locality stored carries out vector quantization with the vector quantization code book of this sound signal type matching to LSF parameter.Finally, the coding parameter of vector quantization is transmitted to multiplexer and is multiplexed in total audio coding frame.
The present invention is based on the LPC coefficient quantization method and apparatus of audio signal classification, being applied at least one coding core uses linear prediction LPC to the odd encoder core encryption algorithm of the coding audio signal process of at least two types, can further improve the quantified precision of the Prediction Parameters of inner linear predictive coding module, thus improve the efficiency of overall digital audio coding algorithms and the subjective sound quality of scrambler.
For AMR-WB+ multinuclear encryption algorithm, apply the many vector quantization code books based on audio classification of the present invention, vector quantization is carried out to wherein LPC coefficient (or LSF parameter of conversion), wherein concrete vector quantization scheme adopts the applying date to be on July 17th, 2012, application number is 201210246780.9, multilevel vector quantization method disclosed in the Chinese patent application that name is called " for voice signal LPC coefficient being carried out to the method and system of multi-stage vector quantization ".
The characteristic of match stop and non-categorical situation below:
(a) complexity
Owing to there is the process to audio signal classification in AMR-WB+, carry out ACELP(voice class signal respectively) and TCX(mixing class signal) coding, therefore the present invention mainly can increase the storage space (about 98k byte) of 1 times, and other complexities are suitable.
(b) performance
Adopt the multilevel vector quantization method in 201210246780.9 " for voice signal LPC coefficient being carried out to the method and system of multi-stage vector quantization ", multiple vector quantization code book of match stop situation and the single vector quantization code book of non-categorical situation, respectively the precision comparison result after the quantification of LPC parameter vector is carried out, as shown in table 1 to table 12 to 12 MPEG typical case cycle testss.
From 12 tables, observe the averaging spectrum distortion representing LPC parameter quantification precision, can think: sorting algorithm is unanimously better than non-categorical algorithm, this also indicates that the many vector quantization code books algorithm based on audio classification can improve code efficiency further.
Table 1 algorithm performance (cycle tests: es01)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.697600 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.502818 | 0.000000 | 0.000000 |
Table 2 algorithm performance (cycle tests: es02)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.662532 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.506807 | 0.000000 | 0.000000 |
Table 3 algorithm performance (cycle tests: es03)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.662490 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.597712 | 0.000000 | 0.000000 |
Table 4 algorithm performance (cycle tests: sc01)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.679964 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.568026 | 0.000000 | 0.000000 |
Table 5 algorithm performance (cycle tests: sc02)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.624548 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.600093 | 0.000000 | 0.000000 |
Table 6 algorithm performance (cycle tests: sc03)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.620681 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.483082 | 0.000000 | 0.000000 |
Table 7 algorithm performance (cycle tests: si01)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.657625 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.530154 | 0.000000 | 0.000000 |
Table 8 algorithm performance (cycle tests: si02)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.735683 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.701430 | 0.000000 | 0.000000 |
Table 9 algorithm performance (cycle tests: si03)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.612262 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.366940 | 0.000000 | 0.000000 |
Table 10 algorithm performance (cycle tests: sm01)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.731752 | 0.000000 | 0.000000 |
| Sorting algorithm | 0.475733 | 0.000000 | 0.000000 |
Table 11 algorithm performance (cycle tests: sm02)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 1.051757 | 0.423729 | 0.000000 |
| Sorting algorithm | 0.800497 | 0.847458 | 0.000000 |
Table 12 algorithm performance (cycle tests: sm03)
| Algorithm | Averaging spectrum distortion (dB) | 2 ~ 4dB ratio (%) | > 4dB ratio (%) |
| Not sorting algorithm | 0.643514 | 0.423729 | 0.000000 |
| Sorting algorithm | 0.626824 | 0.847458 | 0.000000 |
The foregoing is only preferred embodiment of the present invention, not in order to limit the present invention, all any amendments done within the spirit and principles in the present invention, equivalent replacement and improvement etc., all should be included within protection scope of the present invention.
Claims (10)
1., for a LPC coefficient quantization method for audio-frequency signal coding, it is characterized in that, comprise the steps:
S1, determine the type of input audio signal based on predetermined rule;
S2, linear prediction process is performed to input audio signal, calculate LPC coefficient;
S3, for different sound signal types, application and the vector quantization code book of this sound signal type matching carry out vector quantization to described LPC coefficient;
Described method also comprised before step S1:
Signal model for different audio signals type constructs the vector quantization code book mated most with each sound signal type needed for coding respectively and is stored in this locality.
2. method according to claim 1, is characterized in that,
Described step S2 comprises further:
It is the LSF parameter of equivalence by described LPC coefficients conversion;
Described step S3 comprises further:
For different sound signal types, application carries out vector quantization with the vector quantization code book of this sound signal type matching to described LSF parameter.
3. method according to claim 2, is characterized in that, in described step S3, vector quantization adopts multilevel vector quantization method.
4. method according to claim 1, is characterized in that, described method comprises further:
S4, send the coding parameter of vector quantization to multiplexer and be multiplexed in total audio coding frame.
5., for a LPC coefficient quantization device for audio-frequency signal coding, it is characterized in that, comprising:
Memory module, the vector quantization code book mated most with each sound signal type needed for the coding that the signal model for storing for different audio signals type builds respectively;
Audio types determination module, for determining the type of input audio signal based on predetermined rule;
Linear prediction processing module, for performing linear prediction process to input audio signal, calculates LPC coefficient;
Spectrum parameter quantification module, for for different sound signal types, applies and carries out vector quantization with the vector quantization code book of this sound signal type matching to described LPC coefficient.
6. device according to claim 5, is characterized in that, described linear prediction processing module comprises further:
LPC coefficients calculation block, performs linear prediction process for input audio signal, calculates LPC coefficient;
Equivalency transform module, for by described LPC coefficients conversion being the LSF parameter of equivalence.
7. device according to claim 6, is characterized in that, described spectrum parameter quantification module is further used for for different sound signal types, and application carries out vector quantization with the vector quantization code book of this sound signal type matching to described LSF parameter.
8. device according to claim 5, is characterized in that, the coding parameter of vector quantization is sent to multiplexer and is multiplexed in total audio coding frame by described device.
9. an odd encoder core audio coding method, is characterized in that, comprises the steps:
A, the type of input audio signal to be analyzed;
B, for multiple sound signal type, adopt corresponding multiple coding core to encode, wherein, sound signal that at least one coding checks at least two sound signal types performs linear predictive coding;
It is characterized in that,
Described linear predictive coding is quantized LPC coefficient by the method according to any one of claim 1-4.
10. an odd encoder core audio coding apparatus, is characterized in that, comprising:
Audio signal classification processing module, for analyzing the type of input audio signal;
Multiple coding core, for based on sound signal type to corresponding coding audio signal, wherein, sound signal that at least one coding checks at least two sound signal types carries out linear predictive coding;
It is characterized in that,
Described linear predictive coding is quantized LPC coefficient by the device according to any one of claim 5-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310027233.6A CN103137135B (en) | 2013-01-22 | 2013-01-22 | LPC coefficient quantization method and device and multi-coding-core audio coding method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310027233.6A CN103137135B (en) | 2013-01-22 | 2013-01-22 | LPC coefficient quantization method and device and multi-coding-core audio coding method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103137135A CN103137135A (en) | 2013-06-05 |
| CN103137135B true CN103137135B (en) | 2015-05-06 |
Family
ID=48496872
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310027233.6A Active CN103137135B (en) | 2013-01-22 | 2013-01-22 | LPC coefficient quantization method and device and multi-coding-core audio coding method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103137135B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106104682B (en) * | 2014-01-15 | 2020-03-24 | 三星电子株式会社 | Weighting function determination apparatus and method for quantizing linear predictive coding coefficients |
| PL3000110T3 (en) * | 2014-07-28 | 2017-05-31 | Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V. | Selection of one of a first encoding algorithm and a second encoding algorithm using harmonics reduction |
| CN115376532A (en) * | 2021-05-20 | 2022-11-22 | 广州广晟数码技术有限公司 | Audio encoding method, audio decoding method, audio encoding device, audio decoding device, audio encoding equipment and storage medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1287658A (en) * | 1998-10-27 | 2001-03-14 | 松下电器产业株式会社 | CELP voice encoder |
| CN1468427A (en) * | 2000-05-19 | 2004-01-14 | �����ɭ��ϵͳ��˾ | Gains quantization for a clep speech coder |
| CN102779518A (en) * | 2012-07-27 | 2012-11-14 | 深圳广晟信源技术有限公司 | Coding method and system for dual-core coding mode |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR101393298B1 (en) * | 2006-07-08 | 2014-05-12 | 삼성전자주식회사 | Method and Apparatus for Adaptive Encoding/Decoding |
-
2013
- 2013-01-22 CN CN201310027233.6A patent/CN103137135B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1287658A (en) * | 1998-10-27 | 2001-03-14 | 松下电器产业株式会社 | CELP voice encoder |
| CN1468427A (en) * | 2000-05-19 | 2004-01-14 | �����ɭ��ϵͳ��˾ | Gains quantization for a clep speech coder |
| CN102779518A (en) * | 2012-07-27 | 2012-11-14 | 深圳广晟信源技术有限公司 | Coding method and system for dual-core coding mode |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103137135A (en) | 2013-06-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR101565634B1 (en) | APPARATUS FOR ENCODING AND DECODING OF INTEGRATed VOICE AND MUSIC | |
| TWI423252B (en) | Multi-mode audio signal decoder, multi-mode audio signal encoder, methods and computer program using a linear-prediction-coding based noise shaping | |
| JP6214160B2 (en) | Multi-mode audio codec and CELP coding adapted thereto | |
| KR101435893B1 (en) | METHOD AND APPARATUS FOR ENCODING / DECODING AUDIO SIGNAL USING BANDWIDTH EXTENSION METHOD AND Stereo Coding | |
| TWI444990B (en) | Audio encoder, audio decoder and related methods for processing multi-channel audio signals using complex prediction | |
| KR101792712B1 (en) | Low-frequency emphasis for lpc-based coding in frequency domain | |
| EP2673771B1 (en) | Efficient encoding/decoding of audio signals | |
| EP2814028B1 (en) | Audio and speech coding device, audio and speech decoding device, method for coding audio and speech, and method for decoding audio and speech | |
| MX2013009346A (en) | Linear prediction based coding scheme using spectral domain noise shaping. | |
| EP2772912B1 (en) | Audio encoding apparatus, audio decoding apparatus, audio encoding method, and audio decoding method | |
| BRPI0612987A2 (en) | hierarchical coding / decoding device | |
| WO2012053150A1 (en) | Audio encoding device and audio decoding device | |
| KR20100063086A (en) | Temporal masking in audio coding based on spectral dynamics in frequency sub-bands | |
| KR102622804B1 (en) | Backward-compatible integration of harmonic transposer for high frequency reconstruction of audio signals | |
| US9240192B2 (en) | Device and method for efficiently encoding quantization parameters of spectral coefficient coding | |
| CN103137135B (en) | LPC coefficient quantization method and device and multi-coding-core audio coding method and device | |
| KR20090016343A (en) | Apparatus and method for encoding / decoding audio and audio signals using HHT | |
| KR101455648B1 (en) | Method and System to Encode/Decode Audio/Speech Signal for Supporting Interoperability | |
| KR102834523B1 (en) | Backward-compatible integration of harmonic transposer for high frequency reconstruction of audio signals | |
| Setiawan et al. | On the ITU-T G. 729.1 silence compression scheme | |
| Beaugeant | Smart Transcoding between CELP speech codecs through voiced oriented pitch mapping |
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 | ||
| TR01 | Transfer of patent right | ||
| TR01 | Transfer of patent right |
Effective date of registration: 20220510 Address after: 510530 No. 10, Nanxiang 2nd Road, Science City, Luogang District, Guangzhou, Guangdong Patentee after: Guangdong Guangsheng research and Development Institute Co.,Ltd. Address before: 518057 6th floor, software building, No. 9, Gaoxin Zhongyi Road, high tech Zone, Nanshan District, Shenzhen, Guangdong Province Patentee before: SHENZHEN RISING SOURCE TECHNOLOGY Co.,Ltd. |