Source Identification and Separation Using Global Matrix Parameters of ICA

Abstract

Successful separation of independent sources using blind source separation (BSS) techniques requires estimating the number of independent sources in the mixture. Independent component analysis (ICA) is on of the widely used BSS techniques for source separation and identification in audio and bio signal processing. This paper has proposed the use of determinant of the global matrix of ICA as a measure of the number of independent and dependent sources in a mixture of signals. The paper reports experimental verification of the proposed technique where the values of the determinant are seen to be closely based on the number of dependent sources in the mixture.