Fixed-point complex ICA algorithms for the blind separation of sources using their real or imaginary components

Abstract

The complex-valued signal model is useful for several practical applications, yet few algorithms for separating complex linear mixtures exist. This paper develops two algorithms for separating mixtures of independent complex-valued signals in which statistical independence of the real and imaginary components is assumed. The procedures extract sources assuming that the kurtoses of either the real or imaginary components are non-zero. Simulations indicate the efficacy of the methods in performing source separation for wireless communications models. © Springer-Verlag Berlin Heidelberg 2006.