Source amplitude spectral information in matched-field localization.

Abstract

This paper examines a variety of approaches to make use of knowledge of the relative amplitude spectrum of an acoustic source (but no knowledge of the phase spectrum) in multifrequency matched-field processing for source localization. A common example of this procedure involves cases where the source amplitude spectrum can be considered flat over the frequency band of interest. The primary issue is how to combine the information of complex acoustic fields at multiple frequencies, given the unknown phase spectrum. Approaches examined include maximum-likelihood phase estimation, pair-wise processing, and phase rotation to zero the phase at a specific sensor or to zero the mean phase over the array. The performance of the various approaches (processors) is quantified in terms of the probability of localizing the source within an acceptable range-depth region, as computed via Monte Carlo sampling over a large number of random realizations of noise and of environmental parameters. Processor performance is compared as a function of signal-to-noise ratio, number of frequencies, number of sensors, and number of time samples (snapshots) included in the signal averaging.