Quantitative ambiguity analysis for matched-field parameter estimation

Abstract

Matched-field methods concern the estimation of source location and/or ocean environmental parameters by matching the measured signal field with the modeled signal field. The resulted ambiguity output is often characterized by a multimodal structure since the typical signal field is a highly nonlinear function of the embedded parameters. In the high signal-to-noise ratio (SNR) region, the peak of the true parameter protrudes prominently above the noise and can be located accurately; below some SNR, the true peak often falls below the noise level and is obscured by other ambiguous peaks, leading to a larger estimation error and the well-known threshold phenomenon. Therefore, issues on the output ambiguity structure are very important for the development of any matched-field algorithm operated in low SNR scenarios (e.g., passive source localization). To understand some fundamental ambiguity behaviors, a quantitative approach for error analysis is developed in the context of the maximum likelihood estimate. The derived probabilistic error at each scanning parameter point is determined by the signal ambiguity level, the distance to the true parameter point, and the SNR. Accordingly, the threshold SNR can be located by comparing the relative size of the main lobe and side lobe error contributions.