In active sonar systems, a clutter-dominated background is often the limiting factor affecting detection performance. Sources of clutter typically violate the central-limit-theorem conditions that lead to Gaussian distributed bandpass measurements, necessitating more general statistical models to represent their effect on the system. The generalized Pareto distribution (GPD) is a common phenomenological model for clutter, with its shape parameter representing severity through the heaviness of the distribution tail. The focus of this report is on techniques for representing active sonar clutter with the GPD model and assessing the degradation in detection performance as the clutter severity increases. The GPD shape parameter is interpreted through its relationship to the K-distribution shape parameter to understand what values are appropriate in different modeling scenarios (e.g., ranging from mild to extremely heavy-tailed clutter). A comparison of parameter estimators leads to one reliably providing an estimate representative of a physically realizable process. Approximations to the design SNR required to achieve a detector operating-point specification (i.e., the detection threshold term in the sonar equation) for the standard signals in GPD clutter are presented as is the J-divergence detection currency when accounting for thresholding. These simple approximations enable more realistic prediction of active-sonar detection performance by accounting for clutter severity through the GPD model.

Acknowledgments

This report was sponsored by the Office of Naval Research, Code 32, Undersea Signal Processing, through Naval Sea Systems Command contract N00024-21-D-6400 under task order N00024-22-F-8714. The author thanks Dr. J. Gelb (ARL:UT) for reviewing this report.