Bayesian Source Localization via Multistep Focalization in Shallow Water

摘要

In a realistic application of matched field processing for source localization, the ability to localize an acoustic source is strongly affected by the uncertainty of the environment. However, accurate measurements of the environment are extremely difficult to obtain in large regions of the ocean. Two general approaches, referred to here as one-step focalization and multistep focalization, are used to overcome mismatch and accurately estimate source location with limited a priori environmental information. Focalization maximizes the posterior probability density over the unknown source and environmental parameters. One-step focalization directly estimates all parameters at a single step, while multistep focalization reduces the inversion problem to a sequence of smaller subsets that have fewer parameters to optimize at each stage. This greatly improves the efficiency of the optimization and gives the less sensitive parameters more chances to contribute to the objective function. The broadband signals recorded by a vertical line array during a Yellow Sea experiment are used to verify the validity of the approach.