Iterative Stack-Based Detection for Unknown ISI Channels

We propose a turbo-style receiver for detecting convolutionallyencoded data transmitted over multiple unknown dispersive channels. The receiver employs two Bayesian maximum likelihood sequence detectors (BMLSD), each of which uses a stack-based tree search algorithm to generate an estimate of the transmitted data. The detectors exchange soft information about the underlying bits, thereby reducing the number of branch explorations required to detect a transmitted block. The proposed structure is particularly attractive when receiver memory is limited, as the iterative exchange of priors can overcome the erasures typically associated with small stack size in tree search algorithms. Simulation results show that the turbo-BMLSD receiver achieves significant performance gains over the original BMLSD algorithm and approaches the performance of MLSD for a known channel.