Model-based signal processing for passive source localization

We address model-based processing approaches for source depth and range estimation using passive sonar. Two model-based methods have been investigated to date: temporal variations in received amplitude (Jemmott and Culver, 2011) and spectral striation (waveguide invariant) matching (Sell and Culver, 2011). We have shown that both methods provide satisfactory results when the environment is sufficiently well known to produce accurate acoustic model predictions. The methods are incoherent or energy-based and are thus less demanding than matched field processing (MFP). Another model-based source depth estimation method that shows promise is mode-filtering (Premus, 1997). It too is model-based and successful application depends upon sufficient environmental knowledge, and again is distinct and less demanding than MFP. Current research is directed toward comparing these methods in terms of correct classification, as well as robustness to errors in environmental knowledge. Work sponsored by ONR Undersea Signal Processing.