Spatial post-filter estimation for speech enhancement in the specific area using a pair of microphone arrays

Abstract

Suppression methods using the beamformer and the Wiener post-filter by microphone array has been proposed. However, these methods are hard to suppress noise on the environment, such as the desired source is surrounded by noise. We propose the suppression method using a pair of microphone arrays arranged in different positions on the environment, such as the desired source is surrounded by noise. We can get different information of sound by using two microphone arrays. We make the post-filter using its different information of sound. This method enhance the desired source to be the sound source in the specific area. Therefore, we divide the room into nine areas by using two microphone arrays. By estimating the cross spectrum between two microphone arrays of each partition area, we make the post-filter to noise suppression. By the experiment results under actual environment, we confirmed that this method was higher in noise suppression performance than MVDR using one microphone array that is general beamformer.