Non-Cancellation Multistage Kurtosis Maximization with Prewhitening for Blind Source Separation

Abstract

Chi et al. recently proposed two effective non-cancellation multistage (NCMS) blind source separation algorithms, one using the turbo source extraction algorithm (TSEA), called the NCMS-TSEA, and the other using the fast kurtosis maximization algorithm (FKMA), called the NCMS-FKMA. Their computational complexity and performance heavily depend on the dimension of multisensor data, that is, number of sensors. This paper proposes the inclusion of the prewhitening processing in the NCMS-TSEA and NCMS-FKMA prior to source extraction. We come up with four improved algorithms, referred to as the PNCMS-TSEA, the PNCMS-FKMA, the PNCMS-TSEA(p), and the PNCMS-FKMA(p). Compared with the existing NCMS-TSEA and NCMS-FKMA, the former two algorithms perform with significant computational complexity reduction and some performance improvements. The latter two algorithms are generalized counterparts of the former two algorithms with the single source extraction module replaced by a bank of source extraction modules in parallel at each stage. In spite of the same performance of PNCMS-TSEA and PNCMS-TSEA(p) (PNCMS-FKMA and PNCMS-FKMA(p)), the merit of this parallel source extraction structure lies in much shorter processing latency making the PNCMS-TSEA(p) and PNCMS-FKMA(p) well suitable for software and hardware implementations. Some simulation results are presented to verify the efficacy and computational efficiency of the proposed algorithms.