Non-Gaussian Likelihood Detectors for Broadband Active Sonar

In shallow water environments or using broadband processing,detection statistics enerated from sonar data can exhibit highly non-Gaussian statistics due to the discrete nature of the returns from different clutter elements in different range resolution cells. Optimal likelihood detectors rely crucially on accurate estimation of the clutter distribution, and poor fits to an assumed Gaussian probability distribution function may lead to increased false alarm rates when the true PDF is heavy-tailed. We propose to implement a new class of physics-based models of non-Gaussian clutter statistics, developed under phase I of this effort, in order to construct improved likelihood detectors. Our approach smoothly interpolates between the low frequency regime (100 Hz to 1000 Hz) to mid and high frequency ranges. This is because our acoustic models have natural low, middle and high frequency implementations, and because of the adaptive nature of our models for the clutter returns. We also propose to develop criteria to optimize active waveforms for suppression of clutter interference. Finally, we propose to specialize the general development to the waveform and aperture characteristics of various tactical systems to demonstrate improved performance.