Simultaneous extraction of signal using algorithms based on the nonstationarity

Abstract

This article is based on the blind extraction of signals from the observations of a linear and instantaneous mixture. Criteria that use the nonstationarity of the sources are analyzed. One use the algorithm of natural gradient to optimize one of these criteria and its result compares the one of the algorithm of fixed point. In the experiments included, the gradient algorithm allows the simultaneous extraction of the desired number of sources. © Springer-Verlag 2004.