BLIND SEPARATION OF NON-CIRCULAR SOURCES

Jerome Galy, Claude Adnet, Eric Chaumette, Guillaume Gelle 3, claude.adnet@airsys.thomson-csf.com, guillaume.gelle@univ-reims.fr
galy@lirmm.fr,, claude.adnet@airsys.thomson-csf.com, guillaume.gelle@univ-reims.fr

Blind Source Separation is now a well known problem. When a priori informations about the propagation or the geometry of the array are not available, the model can be generalized to a blind source separation model. Il supposes the statistical independence of the sources and their non-gaussianity. In this paper, we focus on an algorithm, called Canonical Correlation Analysis, based on the use of second order statistics.