Broadband acoustic source localization

Abstract

This paper examines aspects of matched field processing for locating a broadband source in deep water. Concentration is on nearby sources (within 5 km of a vertical receiving array), on frequencies from 100 to 300 Hz, and on deterministic, simulation “data” only. The noise field has been modeled as Gaussian, uncorrelated, white noise while the acoustic field has been modeled by the fast field program, with subsequent processing by conventional and maximum likelihood (ML) methods to generate range‐depth ambiguity surfaces (AMSs). The effects of mismatch in the description of sea surface roughness have been examined and it has been found that it can lead to significant errors in peak location on an AMS, particularly at the higher frequencies. However, incoherent frequency averaging of AMSs can, in the presence of relatively small mismatch, overcome the effect of such mismatch, even though almost all the component frequencies show errors in their predictions. Moreover, consideration of range averaging plus frequency averaging can, under such conditions, improve those cases not resolved by frequency averaging alone. It is hypothesized that frequency and/or range averaging will help only when the mismatch is sufficiently small that the AMSs have at least local maxima near the true source location.