A new behavior of higher order blind source separation methods for convolutive mixture

The high order statistics (HOS) based contrasts are widely used in blind sources separation. We propose in this work a new contrast based on cross-trispectra. It presents two main advantages over others traditional contrasts. Firstly, it concerns the computational cost, and secondly the extension takes into consideration the extraction of the independent sources in the presence of more than one gaussian sources. Finally, we shoe experimentally that the sources can be separated in presence of more than one gaussian sources.