A deterministic blind separation approach based on GA for digital communication signals

A simple and effective blind approach is proposed for digital communication signals' separation task. By adopting deterministic contrasts, the general statistical independence assumption hold by conventional blind separation schemes is relaxed. Simplex Genetic Algorithm is applied in optimization procedure to avoid failing into local minima or saddle points, and could easily incorporate priori into fitness function to simplify optimization procedure. Moreover, GA based approach is a general framework and need little modification for different modulation mode, comparing with algebra or gradient based realizations. Presented simulations illustrate its effectiveness both in MIMO separation problem of correlated sources and blind STBC detection scheme.