Model-based recursive Bayesian state estimation for single hydrophone passive sonar localization

This dissertation derives a novel signal processing framework that employs passive sonar to estimate the location of an acoustic source. The approach relies on modeling acoustic propagation, and is designed to estimate the location of a moving source radiating a low frequency tonal signal in a shallow water undersea environment. As the source moves through the ocean, spatial variability in the acoustic field resulting from multi-path propagation is mapped into temporal amplitude modulation of the received signal. The received signal is measured using a single hydrophone on the sea floor. The recorded amplitude modulation contains information about source location that is encoded by propagation effects.;Unfortunately, accurate prediction of the received signal amplitude modulation is complicated by significant uncertainty regarding values of ocean environmental parameters that acoustic propagation models are sensitive to. To increase robustness against the resulting data-model mismatch, the localization framework uses Monte Carlo acoustic propagation models to estimate the acoustic field uncertainty resulting from uncertainty about ocean environmental parameters. The acoustic field uncertainty is represented with probability density functions, which are calculated using parametric, maximum entropy and kernel density estimators. The resulting probability density functions capture prior knowledge about acoustic propagation and environmental parameters without making unnecessary assumptions.;The localization framework uses a grid-based Bayesian approach to recursively calculate the posterior probability density function of source location based on Monte Carlo model results. The Bayesian approach provides a principled method by which prior knowledge about environmental parameters, source dynamics, source characteristics, and acoustic propagation physics can be combined with received acoustic data to calculate the most likely location of a source.;The result is a passive sonar signal processing algorithm that uses acoustic propagation model results to estimate the location of a moving source radiating a tonal signal in shallow water based on the amplitude modulations recorded on a single hydrophone. The localization framework derived in this dissertation is distinct from Bayesian matched field processing in that it neither relies on a vertical array nor computes modal amplitudes from the received data. Results using SWellEx-96 Event S5 data are shown for each step of the framework.