A nonlinear matched-field processor for detection and localization of a quiet source in a noisy shallow-water environment

Abstract

Matched-field processing is a technique that has been developed for the detection and localization of an acoustic source in an underwater environment. Conventional matched-field processing utilizes cross correlations between the complex acoustic pressures measured from the elements of a submerged array of hydrophones and theoretically predicted complex acoustic pressures obtained from an appropriate model of the environment with an assumed source location. Modal matched-field processing has been recently developed for the detection and localization of an acoustic source in a shallow-water waveguide environment, where modal propagation of acoustic energy dominates. Modal matched-field processing utilizes the cross correlations between predicted complex modal amplitudes and complex modal amplitudes calculated from the measured complex pressures. The ambient noise field in a shallow-water environment has a correlated component that propagates to the hydrophone array through the discrete, trapped modes of the waveguide. Cross correlations due to this modal noise compete with signal cross correlations in the cross-spectral matrix used in conventional matched-field processing. In this article, it is shown that modal noise coming from a large number of discrete, randomly distributed, farfield sources contributes insignificantly to the off-diagonal elements of the cross-spectral matrix used in modal matched-field processing. This is to be contrasted with signal cross correlations, which contribute significantly to these off-diagonal elements. This provides a separation of signal from noise that is not available with conventional matched-field processing. In this article, nonlinear (off-diagonal), modal matched-field processing algorithms are demonstrated that exploit this separation of signal and modal noise. The superiority of these off-diagonal algorithms over conventional matched-field processing in a low signal-to-noise situation is demonstrated using computer simulations of a shallow-water Pekeris waveguide.