A fast on line signal separation algorithm

In this paper, a fast RLS-type adaptive algorithm is devised for on-line signal separation. The described method assumes that n sensors are available from which n independent mixed signals are obtained. The proposed algorithm updates the elements of a decoupling system in a manner to minimize output cross correlation functions. This is achieved through the computation of the steepest decent direction of a constrained cost function that describes the output cross correlation functions of the decoupling system, under equal output power constraint. This direction is updated recursively, as new information comes in. Illustrative examples are given.