A Target Sound Extraction via 2ch Microphone Array Using Phase Information

Abstract

In this paper, we propose a novel learning method of linear filters for a target sound extraction in a non—stationary noisy environment via a microphone array with 2 elements. The method is based on a phase difference between two microphones, which is detected from outputs of the Hilbert transformer whose length is corresponding to a fundamental period of vowel parts of speech signals. The cue signal, which has a correlation with a power envelop of target sound, is generated using a mean square of phase difference and applied to the learning. A superior performance of the proposed method is presented by several computer simulation results.