Single Channel Blind Source Separation using the Best Characteristic Basis

This paper proposes a novel technique for separating single channel recording of speech mixture using a hybrid of maximum likelihood and maximum a posteriori estimators. In addition, the new algorithm proposes a new approach that accounts for the time structure of the source signals by encoding them into a set of basis filters that are characteristically the most significant. Real time testing of the new algorithm has been conducted and the obtained results are very encouraging.