Gibbs sampling classification of QAM signals in frequency selective channels

In this paper, we present a method for QAM signal classification at the receiver, when signals are transmitted through frequency selective channels. Utilizing the Gibbs sampler, a classification of the constellation is achieved, along with joint estimation of all unknown system parameters including the data symbols. When the channel parameters are unknown, additional steps are taken for parameter estimation without altering the general approach. The method is shown to perform well on constellations with different cardinalities, resulting in an algorithm with a relatively low computational complexity and fast convergence.