Development of detection device for dugong calls

Abstract

An acoustical approach for research on marine mammals has been a very active research method in recent years. Dugong (Dugong dugon) is one of the highly endangered species, which are strictly-marine herbivorous and mainly inhabit coastal areas. In order to detect dugong calls from recorded data, several algorithms have been adapted by researchers in the analyzing process. However, the number of misses in the detection is still non-zero. The sound of snapping shrimp recorded in a wide range (2-300 kHz) is one of the main background noises that makes the detection of dugong calls difficult in warm shallow waters. Impulse elimination was employed in the system to get rid of the snapping shrimp noise. In order to improve the performance of the detection system by increasing the detection rate and decreasing the number of misses, two new algorithms were tested in the experiment. The experimental results for the new algorithms including impulse elimination and the cepstrum method are presented in this paper.