Microphone array speech enhancement system combining ICA preprocessing in highly noisy environments

In environments of strong background noise and reverberations, enhancement performance of classic microphone array enhancement systems is still not enough. Since ICA (independent component analysis) can separate independent sources only from observations, in this paper ICA is applied to highly noisy sensor signals to obtain cleaner speech signals, which are taken as inputs to the following microphone array enhancement systems. Experiments on real data show that the procedure of ICA preprocessing is effective in suppressing ambient noise, and dramatically improving the quality of input signals. Performance of microphone array speech enhancement system combining ICA preprocessing is greatly promoted.