An Iterative Signal Detection Algorithm Based on Bayesian Belief Propagation Ideas

We investigate the construction of an iterative algorithm for signal detection based on Pearl´s Belief Propagation algorithm. Two main issues arise: the need to modify the graphical model of the problem in order to force the creation of an iterative algorithm, and the need to find suitable approximations to various probability distribution functions so as to allow an efficient implementation of the algorithm. The resulting algorithm is similar to a conventional deflation-based approach but it can make soft rather than hard decisions.