Entropy-based detection and classification of Bryde’s whale vocalizations

Abstract

Investigation of a cetacean species distribution, periodicity, and population is made possible by long-term monitoring of their vocalizations. In this study, a sample entropy (SampEn) method is proposed for the automatic detection of Bryde’s whale calls. Additionally, the k-means approach is presented to automatically classify the whale signals to whale calls and noise depending on the signal-to-noise ratio instead of using a manual threshold approach. The performance of the proposed detection scheme is compared to the traditional dynamic time warping (DTW) algorithm. The detection performance comparison result shows that the proposed SampEn scheme effectively detects Bryde’s whale calls in the presence of ambient noise with a higher accuracy and lower error rate performance compared to the template-based DTW algorithm, achieving 90.73% accuracy and 8.17% error rate.