Sequential approach to blind source separation using second order statistics

A general result on identifiability for the blind source separation problem, based on second order statistics only, is presented. The separation principle using second order statistics is first proposed. This is followed by a discussion on a number of algorithms to separate the sources one by one.