Direct-form adaptive equalization using soft information

The problem of adapting the coefficients of a direct- form adaptive equalizer using soft information is investigated. An algorithm for the adaptation of equalizer coefficients when the transmitted symbol sequence is unknown at the receiver is derived. The resulting adaptation procedure also takes into account prior probabilities of the symbols, which may be available, for instance, in a turbo equalization setting. The starting point of the derivation is the expectation-maximization algorithm, which is then simplified into a purely recursive form with low complexity. An equalizer that uses the adaptation algorithm and a turbo system with this equalizer are derived, and the performance improvement afforded by these is demonstrated both by simulation and by tests with the data from the SPACE08 and KAM11 underwater acoustic communication experiments.