BLIND SIGNAL SEPARATION OF ARBITRARY MIXTURES: ADAPTIVE ALGORITHMS AND STABILITY ANALYSES

Scott C. Douglas and Xiaoan Sun

In this paper, we present novel techniques for blindly sepa­ rating mixtures of instantaneously­mixed sources. The al­ gorithms are able to separate mixtures of arbitrary non­ zero­kurtosis sources without specific knowledge of and with­ out estimating the statistics of the individually­extracted sources. Both orthogonal contrast and equivariant prewhit­ ening/source separation implementations are provided. Lo­ cal stability analyses and simulations of the proposed meth­ ods verify their abilities to separate arbitrary source mix­ tures.