Bayesian localization of acoustic sources with information-theoretic analysis of localization performance

Abstract

Approaches investigated to date for localizing acoustic sources include conventional beamforming, matched field processing, and Bayesian methods [e.g., Pitre and Davis, J. Acoust. Soc. Am., 97, 1995], with recent research revisiting Bayesian methods with focalization and marginalization approaches [Dosso and Wilmut, J. Acoust. Soc. Am., 129, 2011]. Information-theoretic bounds on source localization performance were investigated by Meng and Buck [IEEE Trans. Sig. Proc., 58, 2010] extending earlier work of Buck. The present work investigates direct application of Bayes’ Rule to source localization and information-theoretic quantification and analysis of localization performance, taking as an example the localization of a time-harmonic source in a range-independent shallow-water acoustic waveguide. Signal propagation is represented by normal modes, and additive Gaussian ambient noise is represented by a Kuperman-Ingenito model. The localization performance is quantified by the entropy of the Bayesian posterior pdf of the source location, and an information-theoretic interpretation of this performance measure is presented. Comparisons with matched-field localization performance and extensions of the modeling and localization performance analysis to inhomogeneous media are discussed. [Work supported by ONR.]