Joint source-channel coding over Gaussian channels with intersymbol interference using tree structured vector quantization

The transmission of a vector quantized Markov source over channels with intersymbol interference (ISI) using M-dimensional modulation procedures is considered. The source is quantized using a channel matched tree structured vector quantizer (CM-TSVQ) in which the VQ design is performed with respect to channel error matrices computed by simulation. The CM-TSVQ is combined with a recursive block MAP receiver that provides soft decisions and accounts for the residual redundancy left at the vector quantizer output. In order to reduce the complexity of the decoding, computation in the log-domain of the probability metrics is used. The M-dimensional modulation is implemented as M/2 two-dimensional QAM. Optimum power allocation between QAM signal sets is determined. Results are presented for the case of two stage TSVQ, using 2-QAM, and are compared to the OPTA (optimum performance theoretically attainable) bound for Gaussian channel with and without ISI.