Genetic algorithms for blind maximum-likelihood receivers

The ultimate receiver in a communications system is one that minimizes the bit-error rate (BER) or, equivalently, that maximizes the likelihood function. Unfortunately, a maximum-likelihood (ML) receiver can be prohibitively complex in some cases. For instance, in a blind system, where neither the channel nor any part of the transmitted sequence are known, an ML receiver would have to test all possible transmitted sequences to determine the one that minimizes the BER. In this paper, we derive a likelihood function for blind communications, and we use a genetic algorithm as the optimization strategy, at a reasonable computational cost. The performance of the resulting algorithm can be improved by exploiting structural aspects of the transmitted sequence that are normally neglected by blind techniques, such as the presence of some known symbols or of an error-control code. Simulation results are presented to validate the proposal.