Detection enhancement using multiple time-reversed guide sources in shallow water: Analysis of TREX04 data

Abstract

Detection in a monostatic, broadband, active sonar system in shallow water is degraded by propagation-induced spreading. In the TREX04 experiment, performed south of the Hudson Canyon off the coast of New Jersey, 0.25 sec LFMs were transmitted with 500 Hz bandwidths chosen over a 0.5–3.5 kHz range using the NRL 64 element source-receiver array. The transmissions were then echo-repeated by a distant ship at a range varying between 1–5 km. The echo-repeating ship also transmitted one-way, 1 s, LFMs which were used as guide-source signals. These guide-source signals, which contain environmental information, form the basis of a technique for improving detection without having explicit environmental knowledge. Using an empirical-orthogonal-function representation of the set of monostatic guide-source signals, echoes were convolved with the time-reversed orthogonal functions as part of a filter bank. The result is improved probability of detection of noisy echoes using multiple guide-source signals in comparison with a baseline probability of detection using matched-filtering. In this talk, ROC curve improvement is analyzed as a function of the many parameters, e.g., target depth, source aperture, bandwidth, number of guide sources and their spatial and temporal separation. Comparison with numerical simulations is also made. [Work supported by ONR.]