APPLICATION OF PROBABILITY MODELING AND BAYESIAN INVERSION TO PASSIVE SONAR SIGNAL-PROCESSING .1. CONCEPTS AND FORMULATION

This work considers the benefits of, and methods for the application of probability modeling to passive sonar signal processing. The equal-likelihood probability modeling approach systematically accounts for the combination of data, a priori information, propagation models, array models, and linear and nonlinear models for sources, noise, and environmental descriptors. In Part I, model-consistent, nonstationary, non-Gaussian, equal-likelihood probability models are defined. An optimal nonstationary space-time beamformer and maximum a posteriori probability parameter estimation are derived from equal-likelihood distributions based on autocovariance models employed in conventional matched field processing. Combining information in different frequency bands is seen as an application of a natural data fusion property of stationary Gaussian probability modeling. Part II presents simulation results for the ocean acoustic source localization problem for a shallow water ocean environment under low signal-to-noise conditions. These simulations demonstrate the time evolution of the a posteriori distribution for increased data integration time for single and multiple sources, and illustrate data fusion over a range of frequencies and multiple sensors. © 1995, Acoustical Society of America. All rights reserved.