Blind separation of cyclostationary signals

In this paper we propose a new approach to blind source separation that exploits the cyclostationary nature of the sources typically found in communication applications. A new adaptive algorithm is proposed that simultaneously utilizes Higher Order Statistics (HOS) and cyclic moments. The approach enables exploitation of periodicities embedded in the sources such as the carrier frequency or the symbol rate. It is also presented an analysis to obtain the conditions under which the algorithm converges to the desired solutions.