Noise Robust Talker Localization Based on Weighted Csp Analysis with an Average Speech Spectrum for Microphone Array Steering

This paper proposes a noise robust talker localization method based on weighted CSP (Cross-power Spectrum Phase) analysis with an average speech spectrum as the pre-process of microphone array steering. The proposed method consists of two processes. First, CSP coefficients are weighted by analysis weight coefficients based on an average speech spectrum, which is trained with speech database, in advance. Next, the interference noises are reduced on spatial domain by CSP coefficient subtraction. As a result of evaluation experiments in a real room, we confirmed that the proposed method could provide better talker localization performance than the conventional methods.