Spatial filtering techniques applied to source localization

Abstract

Matched-field processing outputs may be degraded by the presence of multiple sources. To isolate single sources, a spatial filtering algorithm has been developed. Element level data are transformed to beam space, filtered in bearing, then transformed back to element level data which are then used in simulated annealing inversions. The efficacy of the approach was demonstrated by comparing filtered and unfiltered ambiguity surfaces of depth, range, and bearing using a cross-correlation cost function in a simulated multisource environment with equally spaced array elements [Stotts et al., J. Acoust. Soc. Am. 113, 2217 (2003)]. Recently, this method was successfully applied to a real data set where localizations had previously been obtained using beam processing, and ground truth bottom properties have been established. Ambiguity surfaces are exhibited to demonstrate filter performance for real data. Examining the spectrum of singular values associated with the filter shows differences between equally and non-uniformly spaced elements, where a modification is required for frequencies greater than ideal. Energy conservation requirements validate filter output. Filter design details and results for both simulated and real data will be presented. Effectiveness as a function of signal-to-noise ratio, measured relative to an interferer, is also examined.