Two-channel conjugate vector quantizer for noisy channel speech coding

Author

Moriya, Takehiro

Author_Institution

NTT Human Interface Lab., Tokyo, Japan

Volume

10

Issue

5

fYear

1992

fDate

6/1/1992 12:00:00 AM

Firstpage

866

Lastpage

874

Abstract

A two-channel conjugate vector quantizer is proposed in an attempt to reduce quantization distortion for noisy channels. In this quantization, two different codebooks are used. The encoder selects the channel code pair that generates the smallest distortion between the input and the averaged output vectors. These two codebooks are alternately trained by an iterative algorithm which is based on the generalized Lloyd algorithm. Coding experiments show that the proposed scheme has almost the same SNR as a conventional vector quantizer for an error-free channel. On the other hand, it has a significantly higher SNR than the conventional one for a 1% error rate. This scheme also has merits in computational complexity and storage requirements. The scheme is confirmed to be effective for a medium bit-rate speech waveform coder

Keywords

data compression; encoding; speech analysis and processing; telecommunication channels; SNR; averaged output vectors; channel code pair; codebooks; computational complexity; error rate; generalized Lloyd algorithm; input vector; iterative algorithm; noisy channel speech coding; quantization distortion; speech waveform coder; storage requirements; two-channel conjugate vector quantizer; Bit rate; Computational complexity; Ear; Error analysis; Iterative algorithms; Mobile communication; Noise reduction; Protection; Speech coding; Vector quantization;

fLanguage

English

Journal_Title

Selected Areas in Communications, IEEE Journal on

Publisher

ieee

ISSN

0733-8716

Type

jour

DOI

10.1109/49.138991

Filename

138991