A blind source extraction algorithm with a unimodular constraint

Abstract

Blind source separation (BSS) is a method for recovering a set of source signals from the observation of their mixtures without any prior knowledge about the mixing process. On the other hand, a method that recovers only one source signal from the observation is called blind source extraction (BSE). The number of parameters needed to be estimated in BSE is smaller than that in BSS, thereby requiring less computational time. In this paper we show a new BSE algorithm and demonstrate that the algorithm can preserve a signal quality, which is one of the important features for applications, such as speech enhancement. Furthermore, we have showed that a unimodular constraint used in this study can eliminate the indeterminacy in numbering of the sources, which cannot be eliminated in other constraints proposed before.