Dereverberation for reverberation-robust microphone arrays

This paper discusses using dereverberation technologies to improve the robustness of beamforming microphone arrays against reverberation. It is known that, while conventional adaptive beamformers are able to effectively cancel the interference from spatially separate noise sources in the absence of reverberation, their efficacy deteriorates in reverberant rooms. To provide a widely applicable solution to this problem, we propose an adaptive multiple point blind equalizer that can shorten all the room impulse responses between the sound sources and microphones. This algorithm reduces the impact of reverberation on the interference cancellation performance when it is used to pre-process the microphone signals prior to beamforming. Experimental results using real conversation data show that a system combining the proposed equalizer and the adaptive beamformer improves noise reduction performance without degrading the speech quality.