The Using of MVDR Beamformer for Extracting Target Speaker

Human-computer interaction and communication technology require an effective speech technology and MA have been common widely applied in numerous speech applications, due to their perspective performance in enhancing the quality and intelligibility of the captured signals. The utilizing of spatially designed distributed microphone array (MA) to perform beamforming technique with an additive conventional temporal filtering, which allows achieving better speech enhancement while rejecting all the background noise, interference signals or third-party speaker to improve the overall speech quality. Minimum Variance Distortionless Response (MVDR) beamformer is one of the most useful signal processing procedure for extracting the desired talker in noisy mixture of speech and surrounding environments. MVDR own the capability of saving the directional target speaker in a certain location while eliminating all background noise. In this contribution, the author suggested a modified Minimum Variance Distortionless Response (MVDR) beamformer to overcome the drawback of speech distortion, which always available in most of realistic acoustic equipments. The illustrated experiments was conducted by using dual - microphone system in real-life living room and the obtained results confirmed the advantage of the proposed method in reducing the speech distortion to 10.5 dB and increased the overall speech quality from 7.2 to 11.2 dB.