Study on automatic extraction of bottlenose dolphin whistles from the background of ocean noise

Xue Rui-chao1, Niu Fu-qiang1, Yang Yan-ming1, Huang Yue-kun1, Liu Wei2
 
1.Third Institute of Oceanography, State Oceanic Administration, Xiamen 361005, Fujian, China.
2.The Reserve Management Office of Chinese White Dolphins and Amphioxus in Xiamen, Xiamen 361005, Fujian,
China.
 
Abstract: 
Bottlenose dolphins (Tursiops aduncus) use the frequency contour of whistles to identify each other and communicate information. In order to research the characteristics of whistles, as much as possible sound data needed. However, there is very difficult to extraction whistles from strong background of ocean noise. Manual recognition and extraction whistles are not only time-consuming, but also lost effective signals with low signal-to-noise ratio. Methods of the fully automatic detection and extraction of whistles are proposed in this paper. The entire process for noise cancellation to a spectrogram of sound data consists of eight steps. Experimental results show that the total extraction accuracy of the method is 92.5%, and the effective extraction accuracy is 80.5%. This method can meet the requirement of most data processing, and provide a certain technical support for acoustic research of marine mammals.
 
Keywords: 

Dolphin Whistles, Signal Extraction, Noise Cancellation, Acoustic Signal Processing

pages: 

127-130

DOI: 

10.35745/icice2018v2.032

Year: 

2018

Published in: 

2nd International Conference on Information, Communication and Engineering (ICICE 2018)

Full paper download: 
PDF icon 127-130.pdf