OVERDETERMINED BLIND SOURCE SEPARATION: USING MORE SENSORS THAN SOURCE SIGNALS IN A NOISY MIXTURE

Marcel Joho 1 , Heinz Mathis 1 , and Russell H. Lambert 2
joho@isi.ee.ethz.ch, mathis@isi.ee.ethz.ch, russ@pivotaltech.com

This paper addresses the blind source separation problem for the case where more sensors than source signals are available. A noisy­sensor model is assumed. The proposed algorithm com­ prises two stages, where the first stage consists of a principal com­ ponent analysis (PCA) and the second one of an independent com­ ponent analysis (ICA). The purpose of the PCA stage is to increase the input SNR of the succeeding ICA stage and to reduce the sen­ sor dimensionality. The ICA stage is used to separate the remain­ ing mixture into its independent components. A simulation exam­ ple demonstrates the performance of the algorithm proposed.