Dugong chirp type classification based on fundamental contour extraction and hierarchical cluster analysis

Abstract

A steady decrease in the Dugong (Dugong dugon) population has been notices worldwide. Considering the social function and utilization rate, dugong chirp classification has become an important bioacoustics research topic, which may help marine biologists monitor the change of their population, study their behavioral context in different environments and optimise the protection policy. However, the objective classification of dugong chirps has not been adequately addressed in the literature. In this paper, an automatic fundamental contour extraction using the tempo-frequency variables to describe the contours of the fundamental frequency of the chirps is presented. With the automatic fundamental contour extraction, the time–frequency information of dugong chirps can be effectively represented, which gives a good classification performance. Following the principles used for the classification of dolphin whistles, three variables were chosen and performed with hierarchical cluster analysis. Based on these results and further visual examination, five categories of chirps were identified. A simple classification criterion was proposed which could be used in the automatic and effective identification of chirp types. Our findings enrich the known repertoire of dugong chirps, and may be useful in future passive acoustic monitoring studies.