Energy-based Nonstationary Acoustic Source Localization in Correlated Noise Environment

Abstract

The generalized energy-based acoustic source localization problem in a correlated noise environment has been addressed in this paper. At first, a Gaussian distribution was employed to characterize the placement deviations of the sensors. And a fractional Gaussian noise (fGn) statistical model was utilized to characterize the acoustic source and noise correlation. Based on Bayesian estimation theory, a generalized energy-based acoustic source localization scheme was presented, wherein the particle-assisted stochastic search (PASS) algorithm was used to determine the acoustic source position from the observations by sensor arrays. Experiment results were presented to validate the effectiveness of the energy-based acoustic source localization, even in the complicated environment with correlated noise, sensor placement deviations, and non-stationarity characteristics.