Microphone array beamforming owns the capability of suppressing surrounding noise fi eld, third-party talker and other interference while extracting the desired speech component
at specifi ed sound location with perspective results. By using the prior information about impinging angle of target talker, the configured distribution of microphones, the observed properties of
MA signals, the characteristic of considered situation, recording environment, the properties of designed geometry of array, the beamformer can preserve the original speech component and
removing background noise at the same time without speech distortion. Superdirective beamformer utilizes the properties of diffuse noise fi eld for saving the desired speaker with high
directivity index and directivity factor. However, under complex and annoying recording situation, the error of current settings of microphone, the movement of speaker’s head, the microphone
mismatches, the effect of internal noise of acoustic equipment, the overall superdirective bamformer’s performance often corrupted. The unaccepted residual noise or speech distortion degrade the speech quality or other measured perceptual metric listener, which causes seriously uncomfortable with listener. In this article, the authors presented an effective approach to modify the formulation of superdirective beamformer to obtain more robust of speech enhancement in realistic recording environment. The simulated results have shown the improvement in reducing speech distortion, decreasing background noise level and enhancing perceptual metric listener.

Speech enhancement, noise reduction, superdirec- tive beamformer, robust, approach.