Spectral Entropy Based Dolphin Whistle Detection Algorithm and Its Possible Application for Biologically Inspired Communication

Abstract

To manually detect dolphin whistle from the growing sound recordings of passive acoustic monitoring is laborious. In order to solve this problem, two dolphin whistle signal detection methods based on the spectral entropy are analyzed in this paper. The adaptive band-partitioning spectral entropy detection algorithm is improved and a mel-frequency band spectral entropy detection method is proposed for the dolphin whistle detection. The experimental results show that the detection methods could realize automatic detection and find the endpoints. The detection probability could reach more than 90% and the endpoint accuracy could reach more than 60% for both methods. The detection methods have some potential applications in biologically inspired communication based on the whistle duration or time delay modulation. A simple example is described by employing Morse code to modulate the text information into various whistle durations. Pool experiment verifies the feasibility of the Morse code based bionic communication method.