Source Separation Based on Second Order Statistics | an Algebraic Approach

| an algebraic approach Ulf Lindgren Department of Applied Electronics Chalmers University of Technology S-412 96 G oteborg, Sweden [email protected] Alle-Jan van der Veen Department of Electrical Engineering Delft University of Technology 2628 CD Delft, The Netherlands [email protected] Two unknown non-white stochastic sources (e.g. speech signals) are dynamically mixed by an unknown multipath channel and subsequently measured by two sensors. The objective is to construct an inverse lter that separates the two signals, based only on their independence. It is known that, under certain conditions, second-order statistics provide su cient information to identify the lter. In contrast to the usual cost function optimization techniques, we propose an algorithm that computes the lter coe cients algebraically, using linear algebra techniques such as the singular value decomposition.