Blind Adaptive Beamformer Based on Independent Component Analysis for Underwater Passive Acoustic Source Separation

The main problem in underwater passive acoustic (UPA) detection was to separate unknown independent signals that propagate from unknown directions. Few algorithms can separate multiple signals and find their directions when there is no a priori knowledge about signals and their directions. We propose an improved algorithm to resolve this problem, allowing the element weights to be adapted to the element delays, which could possibly increase performance and speed up convergence. It is using the narrowband assumption and adjusting complex weights other than delays. Real radiated noises of ships and computer simulation data are used to test the improved algorithm. The results of the experiments reveal the signal separation performance and convergence rate are superior to independent component analysis blind beamformer. The sampling rate could also be less because phase shifts are used instead of time shifts.