Bayesian framework for three dimensional, near field source localization using spherical harmonics

Abstract

Localization of sound sources relative to a receiver is of great interest in many acoustical applications. Within the near field of a spherical array, spherical wave incidence can be assumed and both direction of arrival and radial distance information can be extrapolated. This work presents a two-tiered Bayesian framework, using spherical harmonics, through which the directions of arrival and radial distance of a single, or multiple sources in the near field relative to a spherical microphone array receiver can be estimated. The first tier estimates the number of sources present in the sound field, while the second tier estimates the direction of arrival and radial distance parameters. Two models for spherical wave incidence on a spherical array, both based on spherical harmonic beamforming techniques, are formulated and analyzed. This work presents analysis of the Bayesian framework and both models to demonstrate the feasibility of model-based Bayesian inference for three dimensional source localization of one or more sources in complex noise environments.