Markov Chain Monte Carlo Detection for Underwater Acoustic Channels

In this work, we develop novel statistical detectors to combat intersymbol interference for frequency selective channels based on Markov Chain Monte Carlo (MCMC) techniques. While the optimal maximum a posteriori (MAP) detector has a complexity that grows exponentially with the constellation size and the memory of the channel, the MCMC detector can achieve near optimal performance with a complexity that grows linearly. This makes the MCMC detector particularly attractive for underwater acoustic channels with long delay spread. We examine the effectiveness of the MCMC detector using actual data collected from underwater experiments. When combined with adaptive least mean square (LMS) channel estimation, the MCMC detector achieves superior performance over the direct adaptation LMS turbo equalizers (LMS-TEQ) for a majority of data sets transmitted over distances from 60 meters to 1000 meters.