A new on-line negentropy-based algorithm for blind source separation

Negentropy is one of the principal techniques for independent component analysis. It serves as a multi-purpose tool for both blind signal separation (BSS) and blind signal extraction (BSE). However, the main and most widely used algorithm based on negentropy, namely Fast-IC works in batch mode. A fast on-line operation with a fast convergence rate is very useful in tracking non-stationary sources. In this paper, we modify the cost function of negentropy to produce an improved on-line algorithm. Simulation results of the proposed algorithm prove its good performance and remarkable convergence rate.