EIGENVALUE PROBLEM APPROACH TO THE BLIND SOURCE SEPARATION - OPTIMIZATION WITH A REFERENCE SIGNAL

The eigenvalue problem approach to the blind source separation [L. Mol gedey and H. G. Schuster, Phys. Rev. Leb. 72, 3634 (1994)] is reinvest igated. The essential assumption is that the source signals should be statistically independent for the eigenvalue method to be applicable. When the source signals are correlated, unfortunately, this elegant ap proach faces a serious problem of optimization. We propose that by emp loying a reference signal in the separation procedure, the reconstruct ed signals that have an optimum minimum mismatch to the original sourc es can be obtained. The role and the criterion in choosing the referen ce signal will be extensively illustrated. Furthermore, the influences of nonzero correlation between different source signals, finite data length, and channel noises on signal separation will also be fully cla rified. [S1063-651X(98)14110-7].