Source detection and localization performance of matched field processors in a real and uncertain, shallow water, ocean environment

Matched field processing (MFP) combines sophisticated acoustic models with signal processing techniques to localize and/or detect an acoustic source in the ocean. However, inevitable inaccuracies in waveguide characteristics, known as mismatch, can severely degrade performance, motivating the recent development of a number of robust processing algorithms. Unfortunately, relatively few inter-comparisons of multiple robust MFP methods have been made with experimental data, leaving unresolved which processors perform best not just in idealized simulations but in the highly complex, ever-changing, real ocean environment for which they were intended. This thesis addresses this deficiency by exploring the real-world challenges of covariance estimation and source motion and then inter-comparing conventional and adaptive MFP algorithms using real at-sea acoustic data. Quantitative performance metrics are used to compare the matched field processors in the presence of environmental and system mismatch, varying input signal-to-noise ratios, and interfering noise sources.