An approach of adaptive blocking matrix based on frequency domain independent component analysis in generalized sidelobe canceller

In this paper, we propose a combined approach of generalized sidelobe cancelling and frequency domain independent component analysis (FD-ICA) for speech enhancement. In conventional generalized sidelobe canceller, the noise reference are often obtained by differencing between microphones, which always causes speech distortion due to leakage in reference signal. We replace the fixed blocking matrix with an adaptive one by applying the FD-ICA method, which can perform a steep null in target direction. Experiment in real environment presents that the enhanced speech obtained by our method has less speech distortion for better noise reference.