Deep learning approach to passive monitoring of the underwater acoustic environment

Abstract

A cost effective approach to remote monitoring of protected areas such as marine reserves, harbors and restricted naval waters, is to use passive sonar to detect, classify, localize, and track marine vessel activity (including small boats and autonomous underwater vehicles). This paper uses data from a single hydrophone mounted above the sea floor to compare a conventional cepstral analysis method with a deep learning approach for monitoring marine vessel activity. Cepstral analysis of the underwater acoustic data enables the time delay between the direct path arrival and the first multipath to be measured, which in turn enables estimation of the instantaneous range of the source (a small boat). However, this conventional method is limited to ranges where the Lloyd's mirror effect (interference pattern formed between the direct and first multipath arrivals) is discernible. It is shown that a Convolutional Neural Network (CNN) operating on cepstrum data with a regression output improves the single sensor ranging performance by considering additional multipath arrivals. To demonstrate the effectiveness of the approach, a surface vessel is ranged using the CNN and the results compared with the conventional method.