Higher order blind source separation using the cyclostationarity property of the signals

Over the last decade, a great number of higher order (HO) blind source separation methods have been developed, aiming at separating several statistically independent sources without any a priori information about the latter. A performance analysis of some of these methods has shown that the latter perform very well in a great number of situations encountered in radiocommunication contexts but fail in separating several Gaussian sources. Besides, their performance degrades in the presence of background noise which is either spatially correlated or non-Gaussian, a situation encountered in the HF bandwidth. In this context, the purpose of this paper is to present and to analyse the performance of a new family of HO blind source separation methods exploiting the potential cyclostationarity of the received sources to overcome most of the limitations of the classical HO blind source separation methods while keeping their good properties.