Optimal Passive Sonar Signal Processing Using the Waveguide Invariant
This dissertation presents optimal signal processing methods and performance analysis for passive, waveguide invariant (WI)-based acoustic source range estimation in shallow water marine environments. The WI, commonly denoted by β, characterizes the range- and frequency-varying channel fading pattern that can be observed in the time-frequency spectrum of hydrophone data. The structure of the fading pattern is governed by the physics of ducted acoustic propagation and can be exploited to estimate source range using a variety of methods; this work focuses on model-based, single-hydrophone techniques for both narrowband (tonal) as well as broadband sources.
Maximum likelihood (ML) estimators are presented for both β and source range for the case of tonal sources. Estimator performance is analyzed for various signal-to-noise ratios (SNRs) and numbers of tones processed in both Pekeris and complex environments using the KRAKEN normal mode program. Acoustic data from the SWellEx-96 experiment is analyzed, and source range is estimated with root-mean-square error (RMSE) under 3% of source range using knowledge of β for the local environment and 6% using an estimate β obtained from an area several kilometers away.
The Cram´er-Rao lower bound (CRLB) on achievable variance of unbiased range and β estimates is derived for the case of a broadband source in an ideal waveguide and is seen to exhibit similar trends as the performance curves for the ML estimators derived for tonal sources. Additionally, an example is provided showing how the framework for derivation of the bounds can be extended to a complex environment modeled after the SWellEx-96 experiment.
Receiver localization can be performed by combining the time-varying WI-based range estimates with knowledge of the source track, and this has a potentially significant application to autonomous underwater vehicle (AUV) navigation. To this end, three receiver localization methods are presented that use either the Doppler effect, WI-based range estimates, or both. Results from Monte Carlo simulations as well as from processing experimental data demonstrate the potential to localize AUVs with an error on the order of a few hundred meters under realistic assumptions regarding source and environmental parameters.
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 United States License.
Rights for Collection: Duke Dissertations