GB/T 22726-2008Specification for multichannel digital audio coding technology (English PDF)

Foreword

Preface VII Introduction VIII

1 Scope

GB/T 22726-2008 is the Chinese national standard on specification for multichannel digital audio coding technology, in the field of information technology. The /T suffix marks it as a recommended standard: it is not compulsory by itself, but it becomes binding as soon as a contract, a tender or a customer specification calls it up - which in practice is how most foreign buyers meet it. It was issued on 22 December 2008 by the General Administration of Quality Supervision, Inspection and Quarantine of the People's Republic of China; Standardization Administration of China, and has been in force since 1 June 2009. Classification: ICS 35.040, CCS L71. This page is published from the official record of the standard held by the Chinese standards administration: the identification, the dates, the classification and the issuing body are taken from there. The clause text, the tables and the numeric limits are in the document itself, which is delivered complete in English translation.

41 Table B. 2 critical band. 8000Hz, long window

43 Table B. 3 critical band. 8000Hz, short window

43 Table B. 4 critical band. 11025Hz, long window

43 Table B. 5 critical band. 11025Hz, short window

44 Table B. 6 critical band. 12000Hz, long window

44 Table B. 7 critical band. 12000Hz, short window

44 Table B. 8 critical band. 16000Hz, long window

45 Table B. 9 critical band. 16000Hz, short window

45 Table B. 10 critical band. 22050Hz, long window

45 Table B. 11 critical band. 22050Hz, short window

46 Table B. 12 critical band. 24000Hz, long window

46 Table B. 13 critical band. 24000Hz, short window

46 Table B. 14 critical band. 32000Hz, long window

47 Table B. 15 critical band. 32000Hz, short window

2 Normative references

The terms in the following documents become the terms of this standard by reference to this standard. All dated references, followed by all Modifications (not including errata content) or revisions do not apply to this standard, however, parties to agreements based on this standard are encouraged to study Is it possible to use the latest version of these files? For undated references, the latest edition applies to this standard.

GB/T 17975.1-2000 Information technology - Generic coding of moving pictures and their associated audio signals - Part 1 IEC 13818-1.1996)

GB/T 4880.2-2000 Language name code Part 2. 3-letter code (eqv

ISO 639-2.1998) Information technology - Eight-bit single-byte coded graphic character set - Part 1. Latin alphabet 1 3 terms and definitions, abbreviations The following terms and definitions, abbreviations apply to this standard.

4 Overview

4.1 coding The main components of the code are shown in Figure 1 and Table 1. Appendix A gives a more detailed description of some of these technologies.

Note. The solid line represents audio data and the dashed line represents control/auxiliary information. Figure 1 coding principle block diagram Table 1 coding Coding module function Transient detection detects if the input PCM sample contains a transient response Variable resolution analysis filter bank The PCM samples of the audio signal of each channel are decomposed into sub-band signals. Time-frequency resolution of the filter bank Determined by the results of transient detection Cross reorganization When there is a transient in the frame, it is used to reorganize the order of the subband samples in order to reduce the total amount needed to transmit them. Number of bits The human ear hearing model calculates the noise masking threshold of the human ear Optional sum/difference encoding converts sub-band samples of left and right channel pairs into sum/difference channel pairs The optional joint intensity coding uses the human ear's high-frequency sound image localization characteristics to intensity encode the high-frequency components of the joint channel. Global bit allocation allocates bit resources to individual quantization units such that their quantization noise power is lower than the masking threshold of the human ear Linear scalar quantization quantizes subband samples within each quantization unit using the quantization step size provided by the global bit allocation Code book selection Grouping the quantization factors based on local statistical features of the quantization factor and selecting the best codebook from the codebook library Out to each group of quantifiers Quantization factor coding uses the codebook to select the selected codebook and its application range to perform Huffman coding on all quantization factors. Multiplexing packs the Huffman code and auxiliary information of all quantization factors into a complete bit stream

4.2 decoding The main components of decoding are shown in Figure 2 and Table 2. Table 2 decoding Decoding module function Multiple demultiplexing Each codeword is unpacked from the bitstream. Since the Huffman code is a prefix code, its decoding and demultiplexing are Completed in the same step Code book selection Decoding each Huffman codebook used to decode the quantization factor from the bitstream and its application range (ap- Logicrange) Quantization factor decoding is used to decode the quantization factor from the bitstream Quantization unit number reconstruction reconstructs the number of quantization units of each transient segment by the codebook application range Inverse quantization decodes the quantization step size of all quantized elements from the code stream and multiplies it by the quantization factor to reconstruct the subband samples Optional Joint Intensity Decoding Reconstructs Subband Samples of Joint Channels from Subband Samples of Source Channel Using Joint Strength Scale Factor The optional sum/difference decoding reconstructs the sub-band samples of the left and right channels from the sub-band samples of the and/or difference channels Inverse cross recombination, when there is a transient in the frame, reversing the cross-recombination of the quantization factor by the encoder Short/temp window function sequence reconstruction For transient frames, depending on the position of the transient and the perfect reconstruction of the MDCT (PerfectReconstruction) Short and temporary window function sequence to be used to reconstruct the frame Variable Resolution Synthesis Filter Bank Reconstructs PCM Audio Samples from Subband Samples

Note. The solid line represents audio data and the dashed line represents control/auxiliary information. Figure 2 decoding principle block diagram

5 Syntactic structure

5.1 function The functions defined for the description of the syntax are shown in Table 3. Table

3 Special function definitions Function definition Max(x,y) returns the maximum value of x and y Min(x,y) returns the minimum of x and y Ceil(x) returns the smallest integer greater than or equal to x ErrorHandling() error handling Exit() exits decoding GetHuffDim(pCodeBook) returns the dimension of the Huffman codebook pCodeBook GetHuffMidTread(pCodeBook) If the codebook index of the Huffman codebook pCodeBook is Mid-treading, return true; no Then, return false GetNumHuffCodes(pCodeBook) Returns the size of each dimension of the Huffman codebook pCodeBook (the number of Huffman codes) HuffDec (pCodeBook) Huffman decoding from the code stream using the Huffman codebook pCodeBook. HuffDec (pCodeBook) k=0; unBits=0; For(n=0;n< nNumCodes;n++) nShift=pCodeBook[k++]; If(nShift >0) unBits=unBits<< nShift; unBits|=Unpack(nShift); If(unBits==pCodeBook[k++]) returnpCodeBook[k]; k++; Among them, nNumCodes is the size of each dimension of the Huffman codebook pCodeBook (the Huffman code) number) HuffDecDiff(pCodeBook) The difference Huffman decoding is performed from the code stream using the Huffman codebook pCodeBook. The specific implementation is as follows: HuffDecDiff(pCodeBook) nDiff=HuffDec(pCodeBook); nIndex=(nIndex+nDiff)%nNumCodes; returnnIndex; Among them, nNumCodes is the size of each dimension of the Huffman codebook pCodeBook (the Huffman code) Number); nIndex is the memory variable of the codebook pCodeBook, initially by ResetHuffIndex(pCodeBook) Initialization Table 3 (continued) Function definition HuffDecRecursive(pCodeBook) Regression Huffman decoding is performed from the code stream using the Huffman codebook pCodeBook. The specific implementation is as follows: HuffDecRecursive(pCodeBook) k=-1; Do k++; nQIndex=HuffDec(pCodeBook); }while(nQIndex==nNumCodes-1); nQIndex=k(nNumCodes-1)+nQIndex; returnnQIndex; Among them, nNumCodes is the size of each dimension of the Huffman codebook pCodeBook (the Huffman code) number) ResetHuffIndex (pCodeBook, nDefault) Reset the memory variable nIndex of the Huffman codebook pCodeBook. ResetHuffIndex(pCodeBook,nDefault) nIndex=nDefault; This memory variable nIndex will be used in the HuffDecDiff(pCodeBook) function. Unpack(X) Unpacks X-bit unsigned numbers from the code stream

5.2 code stream The code stream looks like this. Bit_Stream() While(Unpack(16)==0x7FFF) Frame();

5.3 frames The frame is as follows: Frame() FrameHeader(); For(nCh=0;nCh< nNumNormalCh;nCh++) UnpackWinSequence(); UnpackCodeBooks(); UnpackQIndex(); UnpackQStepIndex(); If(bUseSumDiff==true&& (nCh%2)==1) UnpackSumDff(); If(bUseJIC==true&&nCh >0) UnpackJicScale(); For(nCh=nNumNormalCh;nCh< nNumNormalCh+nNumLfeCh;nCh++) If(nNumBlocksPerFrm==8) nWinTypeCurrent=WIN_LONG_LONG2LONG; nNumCluster=1; anNumBlocksPerFrmPerCluster[0]=1; Else nWinTypeCurrent=WIN_SHORT_SHORT2SHORT; nNumCluster=1; anNumBlocksPerFrmPerCluster[0]=nNumBlocksPerFrm; UnpackCodeBooks(); UnpackQIndex(); UnpackQStepIndex(); UnpackBitPad(); AuxiliaryData();

5.4 frame header The frame header is as follows: FrameHeader() nFrmHeaderType=Unpack(1); If(nFrmHeaderType==0) nNumWord=Unpack(10); Else nNumWord=Unpack(13); nNumBlocksPerFrm=1<< Unpack(2); nSampleRateIndex=Unpack(4); If(nFrmHeaderType==0) nNumNormalCh=Unpack(3)+1; nNumLfeCh=Unpack(1); Else nNumNormalCh=Unpack(6)+1; nNumLfeCh=Unpack(2); bAuxData=Unpack(1); If(nFrmHeaderType==0) If(nNumNormalCh >1) bUseSumDiff=Unpack(1); bUseJIC=Unpack(1); Else bUseSumDiff=0; bUseJIC=0; I...

8 Multiplexing audio streams in MPEGTS

36 Appendix A (informative) Related coding techniques 37 A.

1 Transient analysis 37 A. 2 human ear hearing model 37 A. 3 global bit allocation 37 A.

4 Huffman code code book selection 37 A. 5 and/difference code 39 A. 6 joint strength code

39 Appendix B (Normative Appendix) Schedule 41 for Decoding B.

1 Quantization step size table 41 B.

2 Critical Band Table 43 B.

3 Huffman code table 51 for decoding the length of the transient segment B.

4 Decoding codebook selection and application range Huffman code table 51 B. 5 decoding Huffman code table for quantization step index 53 B.

6 Huffman code table 57 for decoding quantization factor quotient width B.

7 Huffman code table 57 for decoding steady-state quantization factors B.

8 Huffman Code Table 75 for Decoding Transient Quantization Factor Appendix C (Normative Appendix) Multiplexing Audio Streams in MPEGTS 94 C.

1 Stream_ID 94 C.

2 Stream_Type 94 C. 3 DRA Registration Descriptor (DRAregistrationdescriptor) 94 C. 4 DRA Audio Stream Descriptor (DRAaudiostreamdescriptor) 94 C. 5 STD audio buffer size 96 C. 6 byte alignment

96 Appendix D (informative appendix) under the mix 97 D.

18 Table

13 Additional information for channel settings is determined

20 Table

21 Huffman codebook selection for decoding codebook index

22 Huffman codebook selection for decoding quantization factor

23 Number of bits required to decode the number of packed quotients Huffman codebook selection

24 Variables used to decode the quantization factor

25 Huffman Code Book Selection for Decoding Quantization Step Index Table

26 Variables used in decoding and difference coding decisions

27 Decisions for complete no and differential coding

22 Table 28 and difference encoding decision

29 Default Normal Channel Settings

30 Representation of common channel settings

31 Emission sequence of audio data for each channel in an audio frame