Blind separation of instantaneous mixtures of impulsive stable sources

Abstract

This paper introduces a method of blind separation which extracts impulsive sources from their instantaneous mixtures. The "heavy-tailed", or "impulsive" signals are characterized by the nonexistence of the finite second (or higher) order moments. Such signals can be modeled by real-valued symmetric alpha-stable (SaS) processes. A novel blind source separation (BSS) algorithm for extracting "impulsive" source signals from their observed mixtures is presented. This algorithm is based on the minimum dispersion criterion. A new whitening procedure by the normalized covariance matrix is introduced and used as the first step of the algorithm. Some computer simulations are presented illustrating the performances of proposed method.