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.