Model-based signal detection in snapping shrimp noise

In a number of scenarios, detecting the presence or absence of a known signal may be of practical interest. One such example lies in a communication setting, where packet detection is a vital first step to decode transmitted data. The detection problem can be formulated as a binary hypothesis test within the Neyman-Pearson (NP) framework. Our scenario of interest is warm shallow waters, where the sea floor is inhabited by colonies of snapping shrimp. The ambient noise in such a case is impulsive and exhibits memory. We investigate the performance of optimal detectors corresponding to four additive noise models in snapping shrimp noise. In the literature, proposed detectors typically take only the amplitude statistics of the noise process into account. By also considering the memory, we show that there is substantial improvement in detection performance. The detector in the latter case is based on the recently introduced stationary α-sub-Gaussian noise with memory order m (αSGN(m)) model, which effectively characterizes the temporal amplitude statistics of the snapping shrimp noise process.