Bit-estimate based soft decoding for vector quantization over channels with intersymbol interference

Abstract

We introduce a new technique for source coding over noisy channels with intersymbol interference. The focus is on the decoding problem, and we present decoder structures that allow the decoding to be based on soft estimates of the transmitted bits. The new bit-estimate based decoders are optimal for, so called, linear mapping codebooks, and they provide structured low complexity approximations to optimal decoding for general codebooks. We also investigate encoder optimization, and combined source-channel coding design. Numerical simulations demonstrate that the bit-estimate based decoders are able to outperform a two-stage approach that uses Viterbi detection plus table look-up decoding.