Background and Clutter Mixture Distributions for Active Sonar Statistics

摘要: False alarms in active sonar systems arising from physical objects the ocean (e.g., rocks, fish, or seaweed) are often called clutter. A variety of statistical models have been proposed for representing probability false alarm (Pfa) presence clutter, including log-normal, generalized-Pareto, Weibull, and K distributions. However, owing to potential sparseness clutter echoes within analysis window, a mixture distribution comprising one distributions Rayleigh-distributed envelope (i.e., an exponentially distributed intensity) represent diffuse background scattering noise is proposed. Parameter-estimation techniques based on expectation-maximization (EM) algorithm developed mixtures containing aforementioned While standard EM handles log-normal EM-gradient algorithm, which combines with one-step Newton optimization, necessary generalized-Pareto Weibull cases. The -distributed requires development variant exploiting method-of-moments parameter estimation. Evaluation three midfrequency active-sonar data examples, spanning mildly very heavy-tailed Pfa, illustrates that provide better fit than single-component models. As might be expected, inference clutter-source shape shown less biased using model compared when contain both noise.