A localization method for multiple sound sources by using coherence function

Abstract

In the present study, we propose a technique for estimating the vertical and horizontal directions of sound sources using a robot head incorporating two microphones. The microphones have reflectors that work as human's pinnae. We previously proposed a localization method that uses the reflectors to estimate the 3D direction of a single sound source. The present paper deals with the problem of 3D localization of multiple sound sources. Since silent intervals normally occur when a person speaks, such intervals are thought to convey that only one person is talking, even if two or more people are actually speaking. These intervals can be estimated by calculating the correlativity of two audio signals obtained from two microphones. The correlativity is determined by a coherence function. Using the interval for each sound source, we can detect the 3D directions of multiple sound sources. Experimental results demonstrate the validity of the proposed method.