Direction of arrival estimation in practical scenarios using moving standard deviation processing for localization and tracking with acoustic vector sensors

Abstract

Typical direction of arrival (DOA) estimation is done with sensor arrays consisting of a great number of sensors. It is of interest to perform DOA estimation with few sensor locations. Acoustic vector sensors (AVS) provide one option for DOA estimation in applications where few sensor locations are required. The majority of AVS experiments have focused on stationary sources in laboratory environments where the source signature is known and controlled. Experiments in this paper have been conducted with in-air pressure-pressure (pp) AVS to track moving mechanical noise sources (ground vehicles) in real-world environments with various frequency content and signal to noise ratio. A moving standard deviation (MSD) algorithm has been used with AVS to estimate DOA at multiple sites. Azimuthal DOA accuracy has been shown to be dependent on ground-reflected paths for pp AVS with comparison to analytical models. Utilizing multiple AVS sites, in-air pp AVS are demonstrated to localize sources with complex signatures.