Passive acoustic detection of modulated underwater sounds from biological and anthropogenic sources

This paper describes an algorithm for the automatic detection of a particular class of underwater sounds, using a single hydrophone. It is observed that many life-forms, systems or mechanisms emit distinctive acoustic signatures which are characterized by packets of relatively high frequency sound that are repeated at regular, low frequency intervals. These types of sounds are commonly produced by biological (e.g. fishes and invertebrates) and anthropogenic (e.g. scuba diver) sources. The algorithm exploits a simple feature, extracted from the raw hydrophone signal, which enables robust detection even in conditions of severe background noise. In order to demonstrate how the algorithm can be used, trial applications are presented for the detection of two different kinds of underwater sound source. First, the algorithm is applied to the problem of detecting soniferous fish sounds, showing that it is possible to robustly automate the detection of instances of cusk-eel presence in hydrophone recordings, thereby simplifying the arduous task of human monitoring of long sound recordings in marine biological research. Second, the algorithm is applied to the problem of automatic diver detection in a noisy urban estuary, demonstrating its potential for harbor security and fleet protection.