Automatic detection of marine mammals using spectral entropy

• Published in: The Journal of the Acoustical Society of America Vol. 126; no. 4_Supplement; p. 2230
• Main Author: Erbe, Christine
• Format: Journal Article
• Language: English
• Published: 01-10-2009
• Subjects: Marine; Mysticetes; Odontoceti; Pinnipedia; PN, Arctic
• ISSN: 0001-4966; 1520-8524
• DOI: 10.1121/1.3248998

An automatic detector for tonal and narrow-band signals of a variety of co-habiting marine mammal species is presented. The detector was developed to automatically identify signals in years of autonomous underwater recordings made in the Arctic. These recordings contained calls of odontocetes, mysticetes, and pinnipeds. Three types of automatic detectors were tested; two were based on peak-energy detection, and one was based on spectral entropy computation. A set of test data was assembled by active listening and visual discrimination of spectrograms. Performances of the automatic detectors were compared by means of receiver operating characteristics. The entropy detector outperformed the other two. It ran considerably faster than real time and can be used as a first step in automatic signal analysis, quickly scanning recordings for potential signals. It should be followed by automatic classification, recognition, and identification algorithms to group and identify signals.