Sixth European Conference on Speech Communication and Technology

Budapest, Hungary
September 5-9, 1999

A Novel Channel Distortion Measure for Vector Quantization and a Fuzzy Model for Codebook Index Assignment

J. S. Pan (1), C. S. Shieh (1), T. F. Chiang (2)

(1) Department of Electronic Engineering, National Kaohsiung Institute of Technology, Taiwan
(2) Department of Electrical Engineering, Nan-Tai Institute of Technology, Taiwan

Vector quantization is very efficient for data compression of speech and image. The channel distortions are introduced due to channel noise. Assigning suitable indices to codevectors can reduce distortion due to an imperfect channel. Several codebook index assignment algorithms were proposed. Unfortunately, no algorithm is always better than the others for any bit error rate due to these algorithms are operated under the assumption of some fixed channel bit error rate which is not realistic. In this paper, a novel channel distortion measure is proposed by computing the expected chanel distortion using Belta distribution function. All codebook index assignment algorithms can be optimized based on this distortion measure. Besides, a fuzzy channel optimized vector quantization for codebook design and index assignment is also derived in this paper.

Full Paper (PDF)   Gnu-Zipped Postscript

Bibliographic reference.  Pan, J. S. / Shieh, C. S. / Chiang, T. F. (1999): "A novel channel distortion measure for vector quantization and a fuzzy model for codebook index assignment", In EUROSPEECH'99, 1747-1750.