Blind source separation for acoustic signals using subspace method and frequency domain Infomax

摘要

In this paper, two array signal processing techniques are introduced to the blind separation of acoustic signals to enhance the signal separation performance of the independent component analysis (ICA). The first technique is the subspace method which reduces the effect of room reflection when the system is used in a room. Room reflection is one of the biggest problem in blind signal separation (BSS) in acoustic environment. The second one is the method of solving permutation called IFC (Inter-Frequency Coherency). In this method, a physical property of the mixing matrix, i.e., the coherency in the adjacent frequencies is utilized to solve the permutation. The experiments in a meeting room show that the subspace method improved the score of the automatic speech recognition by 18% and the method of solving permutation reduced the error in the automatic speech recognition to 4%.