Some issues on vector quantization for noisy channels

N. Farvardin, V. Vaishampayan

Research output: Contribution to conferencePaperpeer-review

1 Scopus citations

Abstract

Summary form only given, as follows. Vector quantization is an effective method for source coding which in certain practical applications results in substantial performance improvements compared to other source encoding schemes. In a practical situation the outputs of the vector quantizer are transmitted via a noisy communication channel. As in scalar quantization, an optimal design of the vector quantizer for the noisy channel results in noticeable performance improvements when compared to a conventional vector quantizer designed for a noiseless channel. An algorithm for is developed for vector quantizer design over a noisy memoryless vector channel. The problem is formulated in terms of an encoder/decoder pair where the encoder maps an L-dimensional vector X to a binary codeword and the decoder maps the received binary sequence into an L-dimensional vector Y from a codebook of reconstruction vectors {y 1, y2, ..., ym}. The algorithm (a generalization of the Lloyd algorithm) yields a locally optimum encoder/decoder pair. Extensive numerical results for a first-order Gauss-Markov and Laplace-Markov source and a binary symmetric channel were obtained. The important issue of encoder complexity for the channel-optimized vector quantizer is addressed.

Original languageEnglish
Pages163
Number of pages1
StatePublished - 1988

Fingerprint

Dive into the research topics of 'Some issues on vector quantization for noisy channels'. Together they form a unique fingerprint.

Cite this