Detection and classification of marine mammal clicks

Abstract

Detection of marine mammals within an influence zone of episodal anthropogenic noise source is critical to insure the safety of the animals. Marine mammal clicks are closely modeled by AM/FM signals. The Teager-Kaiser energy operator followed by a threshold detector provides an effective means of detecting AM/FM signals. Classification of the species generating the click is done by finding the maximum cross-covariance between the power spectral density (PSD) of the received click and the PSDs of clicks which have been identified. The classification algorithm will create a new library entry when the cross-covariance is below a predefined threshold. Five species of marine mammals were classified using data from the 3rd International Workshop on Detection, Classification and Localization of Marine Mammals (DCL). This paper presented a computationally-inexpensive marine mammal detection and classification algorithm with high probabilities of detection and correct classification.