Experimental performance analysis of sound source detection with SRP PHAT-β

Abstract

Sound source detection with distributed microphone systems find applications in acoustic imaging, sound source tracking, and beamforming. Sound source location using the steered response power (SRP) algorithm and the phase transform (PHAT) has been shown to have a robust performance in noisy and reverberant environments. Recent works based on a simulation study have reported on further improvements in detection performance with a partial whitening transformation, referred to as PHAT-beta. This paper experimentally analyzes the performance of PHAT-beta in conjunction with steered response coherent power (SRCP) algorithm for sound source detection. Experimental performance results are presented for an 8 microphone perimeter array in different room reverberation and SNR levels. Values for partial whitening parameter (beta) that optimize detection performance under different noise conditions are determined using a receiver operator characteristics (ROC) analysis. Results support conclusions of the simulation study that partial whitening outperforms total (PHAT) and no whitening over a range of signal types and room conditions.