A learning framework for blind source separation using generalized Eigenvalues

Abstract

This paper presents a learning framework for blind source separation (BSS), in which the BSS is formulated as generalized Eigenvalue (GE) problem. Compared to the typical information-theoretical approaches, this new one has at least two merits: (1) the unknown unmixing matrix directly works out from the GE equation without time-consuming iterative learning; (2) The correctness of the solution is guaranteed. We give out a general learning procedure under this framework. The computer simulation shows validity of our method. © Springer-Verlag Berlin Heidelberg 2005.