Preprocessor based on suprathreshold stochastic resonance for improved bearing estimation in shallow ocean

Abstract

Localization of acoustic sources in the ocean is a problem of tremendous interest in underwater acoustics. One of the many factors that limit the performance of processors used for underwater acoustic source localization is the low signal-to-noise ratio (SNR) in the ocean. Preprocessors based on wavelet denoising and suprathreshold stochastic resonance (SSR) have been proposed in the literature for enhancing SNR and thereby improving the performance of processors used for bearing estimation. Denoising techniques based on SSR exploit the fact that the environmental noise in shallow ocean has a heavy -tailed non- Gaussian distribution. In this paper, a method for designing an SSR based preprocessor is presented. It is shown that the use of this preprocessor leads to a significant improvement in the bearing-estimation performance of Bartlett, Multiple Signal Classification (MUSIC) and Subspace Intersection Method (SIM) processors at low SNR. The improved performance appears in the form of a sharper peak in the ambiguity function, lower bias and lower RMS error in bearing estimation, and better resolution of closely spaced sources.