Bayesian Direction of Arrival Estimations for Sound Sources Using a Spherical Microphone Array

A common problem in acoustical applications is the determination of directions of arrival (DoAs) of sound at a receiver. This work aims to address this problem in situations involving potentially multiple simultaneous sound sources by use of spatial filtering, or beamforming, algorithms with a spherical microphone array. This presents a two-level inferential problem of first determining the number of sound sources and then their locations. The solution under investigation here is a probabilistic model-based method, leveraging Bayesian analysis to match analytic models to experimental data. In this method, a large number of models are simulated, predicting the sound field created by various numbers of sources at di erent positions around the receiver. These predictions are then evaluated against the measured data in order to select the simplest such model that can adequately match observations, thereby estimating first the number of sources, then their DoA information. This paper discusses the impact of microphone array configurations upon DoA resolution. It additionally estimates performance of the method based on experimental results for simultaneous multiple sound sources, particularly in noisy or reverberant environments.