MPEG-4 Multi-Layer Bit-Sliced Arithmetic Audio Coding 報告學生 : 游政勳 指導教授 : 尤信程 90. 12. 27.

MPEG-4 Multi-Layer Bit-Sliced Arithmetic Audio Coding

報告學生 : 游政勳指導教授 : 尤信程90. 12. 27

Outline

1. Basic structure of Mutli-Layer Bit-Sliced Scalable

Audio Coding

2. MPEG-4 Bit-Sliced Arithmetic Coding(BSAC)

3. Conclusion

Basic structure of Mutli-Layer Bit-Sliced Scalable Audio Coding(1)1. In order to reach fine-step scalability of audio coding2. Basically design the 16(base layer)-24-32-40-48 -56-64 kbps scalable coder. But the bit-rate of the enhancement layer can be chosen between 1 kbps and a few dozen kbps with any resolution within overall bit-rate,easily ………..bit-rate scalability3. For each layer, only limited frequency band will be coded (ex: the bandwidth of base layer is 3.5 kHz) ………..bandwidth scalability

Basic structure of Mutli-Layer Bit-Sliced Scalable Audio Coding(2)4. The proposed encoder is simply version of AAC,and

additional decoding process is not used in encoder

Basic structure of Mutli-Layer Bit-Sliced Scalable Audio Coding(3)

The priority of element to be encoded is determined according

to the relative importance in following figure

Basic structure of Mutli-Layer Bit-Sliced Scalable Audio Coding(4)

The rearrangement of the bit-patterns of the quantized

spectral data by bit-slice, and four-dimension vectors are

noiseless coded using arthmetic coding with fixed mode

MPEG-4 Bit-Sliced Arithmetic Coding(1)

1. Small step scalability of bit-sliced arithmetic coding

base layer: 16k bps

each enhancemeny layer: 1 kbps

2. In BSAC, the processing unit is coding band that

contain 32 spectral coefficients and use the same

arithmetic coding model index

3. Four dimension bit-sliced vector is divided into two

sub-vectors(sub-vector-0 and sub-vector-1)

MPEG-4 Bit-Sliced Arithmetic Coding(2)

MPEG-4 Bit-Sliced Arithmetic Coding(3)

MPEG-4 Bit-Sliced Arithmetic Coding(4)

MPEG-4 Bit-Sliced Arithmetic Coding(5)4. Information about stereo coding and PNS are

coded by different arithmetic models

5. Scale-factor is differentially coded related to fixed offset and then arithmetic coded using several arithmetic model

6. Coding scheme for the arithmetic model index of

the coding band is the same as scale-factor,but

use different arithmetic model

Conclusion

1. Arithmetic coding is more efficient than Huffman

coding

2. Because of subband audio coding, scalability is

easilily implemented