Full-field likelihood ratio detection for multistatic sensors in uncertain environments

Abstract

As small networked sonobuoys become cheaper and more viable in the submarine warfare community, there is a growing demand for techniques to automatically process detection data from dense fields containing O(100) or more sonobuoys. The presentation describes a Bayesian framework for the incoherent integration of automatic detections from these dense sonobuoy fields. This framework recursively integrates likelihood ratio functions to form a cumulative posterior over a target state space. To be most effective, these methods exploit either empirical sensor performance models or physics-based signal excess sensor performance models. The presentation describes the use of each type of performance model and presents one method for handling uncertainties associated with signal excess predictions. The method presented starts with a prior on a set of possible environmental conditions. When a target is present in the detection data, the cumulative likelihood ratio will peak in the environment most consistent with the true environment.