Two pairwise iterative schemes for high dimensional blind source separation

Abstract

In this paper, we address the problem of scalability to higher dimensional space in blind source separation (BSS), where the number of sources is greater than two. Herein, we propose two schemes of pairwise non-parametric independent component analysis (ICA) algorithms based on Convex Cauchy–Schwarz Divergence (CCS–DIV) for high dimensional problem in BSS. We extend the pairwise method to the scenario of more than two sources, two improved ICA algorithms are developed. Moreover, we employ adaptive sampling technique that samples the signal into small time blocks to evaluate the integration of the CCS–DIV and reduce the computational complexity. The two presented methods enable fast and efficient demixing of sources in real-world high dimensional source applications. Finally, we demonstrate with simulations including a wide variety of source distributions, showing that our presented methods outperform many of the presently known methods in terms of performance and computational complexity.