Localization of multiple environmental sound sources by MUSIC method with weighted histogram

This paper presents an estimation method of environmental sound sources localization based on MUSIC method by small-sized microphone array. MUSIC spectrum depends on array manifold vector calculated from microphone array configuration and eigenvectors provided an observed signal. By considering the relation between a microphone array and MUSIC spectrum, we examine the arrangement of microphones for high accurate localization with fewer number of elements. And to estimate the environmental sound sources, which have various frequency properties, we propose a weighted histogram method for direction of arrival (DOA) estimation considering frequency properties of the observed signal. We introduce histogram obtained by detected peaks by MUSIC spectrum at each frequency bin. In addition, the sound source positions are estimated on intersection of the estimated directions which are obtained by a pair of microphone arrays. In the numerical simulation, the localization accuracy can be improved with reduced aliasing using a microphone array of irregular interspace. And, the proposed method can be improved the estimation accuracy for the electronic sound, which has sparse frequency property.