Amplitude-adaptive vector quantization system

Details Amplitude-adaptive vector quantization system Amplitude-adaptive vector quantization system

1991-09-18

1994-03-15

Chin, Stephen

Pulse or digital communications

Repeaters

Testing

375122, 348417, H04B 1406

Patent

active

RE0345628

DESCRIPTION:

BRIEF SUMMARY
BACKGROUND OF THE INVENTION

1. Field of the Art
This invention relates to a high-efficiency signal coding and decoding system and, particularly, to an amplitude-adaptive vector quantization system.
2. Background Technology
FIGS. 3 and 4 are block diagrams showing the coding and decoding sections, respectively, of a vector quantization system which uses the inner product for the computation of distortion. In FIG. 3, indicated by 1 is the input signal vector derived from a series of input signals divided into blocks in K-samples, 2 is a mean value separation circuit, 11 is the separated mean value, 13 is the mean separated input vector, 3 is a mean value DPCM (Differential PCM) quantizer, 29 is a vector quantizer, 31 is the amplitude gain produced through coding by the vector quantizer, 7 is an amplitude-gain DPCM quantizer, and 33 is the output vector index produced from the code of the vector quantizer. In FIG. 4, indicated by 19 is a mean value DPCM decoder, 20 is an amplitude-gain DPCM decoder, 24 is the DPCM-decoded mean value, 36 is the DPCM-decoded amplitude gain, 35 is a vector quantizing decoder, 38 is the decoded mean value separation vector, 23 is a mean value adding circuit, and 39 is the output signal vector.
Next, the operation wil be described. In the coding section of FIG. 3, the input signal vector 1 which is a series of input signals provided in the form of blocks is received by the mean value separation circuit 2 which separates the mean separated input vector 13 from the intra-vector mean value 11. The separated mean value 11 is subjected to predictive differential quantization (DPCM quantization) by the mean value DPCM quantizer 3, and information of the separated mean value 11 is reduced. The mean separated input vector 13 is subjected to inner product computation by the vector quantizer 29 with the output vector read out of a code book which is a set of output vectors prepared in advance in a statistical manner so that an output vector yielding a maximum inner product is selected. The vector quantizer 29 produces the output vector index 33 for the input mean separated input vector and also the inner product value as the amplitude gain 31 of the input mean separated input vector 13.
Vector quantization is expressed in terms of input signal vectors 13 S=S.sub.1 S.sub.2, . . . , S.sub.K, mean value .mu., amplitude gain .sigma.*, and output vectors y.sub.i =[y.sub.i1, y.sub.i2, . . . , y.sub.ik ], as follows. ##EQU1## where .sigma.=.vertline.S-.mu..UU.vertline. and .vertline.yi=1 are assumed.
Using the inner product, the vector distortion computation is reduced to the form of a product-sum which is easier for execution with a DSP (digital signal processor), and the amplitude component can be obtained at the same time. The amplitude gain 31 produced by the vector quantizer 29 is subjected to DPCM quantization by the DPCM quantizer 7 in the same way as for the mean value 11.
The DPCM-quantized mean value 12, amplitude gain 32 and output vector index 33 obtained as described above are coded and then transmitted or recorded.
The decoding section of FIG. 4 implements decoding for the three components, namely, the DPCM-quantized mean value 12, DPCM-quantized amplitude gain 32 and output vector index 33 provided by the coding section. The DPCM-quantized mean value 12 is DPCM-decoded by the DPCM decoder 19, which yields the decoded mean value 24. Similarly, the DPCM-quantized amplitude gain 32 is DPCM-decoded by the amplitude gain DPCM decoder 20, which yields the decoded amplitude gain 36.
The vector quantization decoder 35 outputs the normalized output vector 37 by decoding the output vector index 33, the amplitude reproduction circuit 22 reproduces the amplitude of the vector as output vector 38 from the decoded amplitude gain 36, the mean value adding circuit 23 adds the decoded mean value 24 to the decoded mean separated output vector 38, and the output signal vector 39 is reproduced.
In the conventional vector quantization system using the inner product for the computation of distortion, th

REFERENCES:
patent: 4558350 (1985-12-01), Murakami
patent: 4560799 (1985-12-01), Spivack et al.
patent: 4639778 (1987-01-01), Yamaguchi et al.
patent: 4670851 (1987-06-01), Murakami
patent: 4680797 (1987-07-01), Renke
patent: 4710812 (1987-12-01), Murakami et al.
patent: 4724535 (1988-02-01), Ono
patent: 4725885 (1988-02-01), Gonzales et al.
patent: 4769826 (1988-09-01), Kubo et al.
Murakami et al "Vector Quantiser of Video Signals" Sep. 29, 1982 two pages.
Murakami et al "Dynamic Multistage Vector Quantization of Images" Electronics and Communications in Japan of Part 1, vol. 69 1986 pp. 93-101.
Gray "Vector Quantization", IEEE ASSP Magazine, Apr. 1984 pp. 4-29.

Affiliated with

Also associated with

Rating

Amplitude-adaptive vector quantization system does not yet have a rating. At this time, there are no reviews or comments for this patent.

If you have personal experience with Amplitude-adaptive vector quantization system, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Amplitude-adaptive vector quantization system will most certainly appreciate the feedback.

Rate now

Comments { 0 }

Profile ID: LFUS-PAI-O-1529328