Statistics of biological noise and performance of generalized energy detectors for passive detection

Abstract

Underwater noise due to snapping shrimp is highly impulsive, and often dominates the ambient noise environment of warm, shallow waters at frequencies above 1 kHz. We report here on the statistics of bandpass snapping shrimp noise data, and on the modeling of the joint distribution of the in-phase and quadrature components using bivariate versions of the generalized Gaussian (GG), generalized Cauchy, and Gaussian-Gaussian mixture models. We evaluate the performance of several generalized energy detectors for passive bandpass detection, by inserting stochastic signals into the noise data. Detection thresholds were measured for an integration time of 0.5 s and false alarm probabilities down to 1%. The locally optimum detector based on the mixture model gave the best weak signal detection performance, with an 8 dB reduction in detection threshold over conventional energy detection. A significance test detector based on the GG model performed 1-2 dB worse, but exhibited better strong signal performance.