Joint Bayesian model selection and blind equalization of ISI channels

Abstract

We consider the problem of joint model selection and blind equalization of inter-symbol interference (ISI) channels. We adopt a Bayesian approach where nuisance parameters are considered random and integrated out. An efficient Markov chain Monte Carlo (MCMC) method is presented to perform Bayesian computation. The proposed algorithm overcomes the problem of delay ambiguity encountered by most existing MCMC algorithms. A simple Metropolis step is employed for model selection circumventing the need for reversible jump Markov chain Monte Carlo (RJMCMC). To the best of our knowledge, the problem of model selection in ISI channels is solved for the first time. The convergence behavior and the Bit Error Rate (BER) performance of our algorithm are demonstrated through computer simulations.