ABSTRACT
Signal processing can simply be defined as a technique or set of techniques to extract the useful information from noisy measurement data while rejecting the extraneous. These techniques can range from simple, non-physical representations of the measurement data such as the Fourier or wavelet transforms to parametric black-box models used for data prediction to lumped mathematical physical representations usually characterized by ordinary differential equations to full physical partial differential equation models capturing the critical details of wave propagation in a complex medium. The determination of which approach is the most appropriate is usually based on how severely contaminated the measurements are with noise and underlying uncertainties. If the signal-to-noise (SNR) of the measurements is high, then simple non-physical techniques can be used to extract the desired information. However, if the SNR is extremely low and/or the propagation medium is uncertain, then more of the underlying propagation physics must be incorporated somehow into the processor to extract the information. Model-based signal processing is an approach that incorporates propagation, measurement, and noise/uncertainty models into the processor to extract the required signal information while rejecting the extraneous data even in
highly uncertain environments — like the ocean. This chapter outlines the motivation and development of model-based processors (MBP) for ocean acoustic applications. We discuss the generic development of MBP schemes and then concentrate specifically on those designed for application in the hostile ocean environment. Once the MBP is characterized, we then discuss a set of ocean acoustic applications demonstrating this approach.
