Locally Optimal Detection of Signals in Underwater Acoustic Noise with Student’s t-distribution

Abstract

Signal detection is imperative in underwater signal processing and digital communication, and based on a knowledge of noise statistics, optimum signal detection in underwater acoustic noise (UWAN) can be more effectively realised. The hypothesis of normal (Gaussian) noise allows the use of matched filter (MF) detectors; accordingly, a locally optimal detector (LO) is designed in this study to improve detection probability (PD) based on the knowledge of noise probability density function. The underwater noise used for validation is real data collected from the sea using broadband hydrophones at the beach of Desaru on the eastern seashore of Johor, Malaysia. The performance of the LO detector is then compared with a conventional MF detector and these are evaluated according to their PD values. For a time-varying signal, a false alarm probability specified as 0.01, and a PD value of 90%, the energy-to-noise ratios (ENR) of the LO are better than those of the MF by 4.2 dB and for fixed frequency signals, the LO is better than the MF by 5.2 dB.