Blind Separation of Convolved Mixtures: A Feed-forward Approach with Whitening

摘要

Blind separation of independent sources from their convolutive mixtures has problems in many real world applications. Especially the feed-forward structure has a side effect, a temporal whitening of the output signal. In this paper we present a feed-forward network architecture without temporally whitening of the estimated output sources. Experiments using speech signals are presented to illustrate the proposed algorithm.