Joint dereverberation and noise reduction using a two-stage beamforming approach

In this paper a two-stage beamforming approach is presented for dereverberation and noise reduction. The first stage comprises a delay-and-sum (DS) beamformer that generates a reference signal that contains a spatially filtered version of the desired speech and interference. In general, the desired speech component at the output of the DS beamformer contains less reverberation compared to reverberant speech signal received at the microphones. The second stage uses the filtered microphone signals and the noisy reference signal to estimate the desired speech component at the output of the DS beamformer. A major advantage over classical approaches is that the proposed approach is able to dereverberate the received desired signal with very low speech distortion. The dereverberation and noise reduction performance is evaluated for a circular microphone array.