Blind separation combined frequency invariant beamforming and ICA for far-field broadband acoustic signals

Abstract

Many famous blind source separation (BSS) in frequency domain have been developed while they can still not avoid the permutation problem. We propose a new BSS approach for far-field broadband acoustic signals via combining the frequency invariant bemforming (FIB) technique and complex-valued independent component analysis (ICA). Compared with other frequency methods, our method can avoid the permutation problem and has much faster convergency rate. We also present a new performance measure to evaluate the separation. Finally, the simulation is given to verify the efficiency of the proposed method. © Springer-Verlag Berlin Heidelberg 2005.