中文 |

Research Progress

China Scientists Propose First Effective Automatic Identification Algorithm of Yangtze Finless Porpoise

Dec 18, 2015

The Yangtze finless porpoise (Image from Internet) 

 Marine mammals are considered to be the most endangered mammals all over the world, among which the river cetaceans (dolphins and porpoises) are the rarest-of-rare ones. River cetaceans inhabit in the restricted habitats heavily populated by humans, and are particularly vulnerable to the humans.

Finless porpoise populations are adversely impacted by human activities that cause water pollution, lack of food and denial of migratory paths. The population size has been declining, and the distribution ranges have been reduced sharply in the past thirty years. Proper survey and management of their populations has therefore become necessary.

Yangtze finless porpoise is one of the most endangered mammals in the world, so it is of great practical significance for surveying and protecting the Yangtze finless porpoise in the wild. In recent years, several Yangtze finless porpoise research surveys have been organized by the Chinese government.

Passive acoustic is becoming a frequently used tool in monitoring marine mammal surveys for study of behavior, migration monitoring. However, many passive acoustic data still need a lot of subsequent processing because these passive acoustic methods cannot detect finless porpoise automatically. The processing of acoustic data by humans is time-consuming and labor intensive. Also the results depend on the experience of the human operators. The acoustic results always need to compare to the visual results.

Researchers from the Institute of Acoustics of the Chinese Academy of Sciences and Tongling River Dolphin National Natural Reserve work together to have proposed an effective automatic identification algorithm of finless porpoise signal for the first time. The new automatic recognition method of finless porpoise has great practical significance for surveying and protecting the Yangtze finless porpoise in the wild.

Researchers propose an effective recognition characteristics vector of Yangtze finless porpoise acoustic signals firstly. The new method combines Hilbert marginal spectrum and Fourier transform to extract finless porpoise. The identification result indicates that the 10-D recognition vector feature vector F can well express the time-frequency characteristics of the finless porpoise signal. As well, the recognition system based on BP artificial neural network can accurately identify the finless porpoise signal. 

Some experimental acoustic data files of finless porpoise are used to test the validity of the automatic identification algorithm. 238 finless porpoise acoustic signals are detected. The correct identification probability of the algorithm proposed reaches 93%, according to the human observation on the time-frequency spectrum.

Research has mainly been focused on the fundamental high frequency from Fourier transform. And the usage of low frequency components from time-frequency spectrum has not been fully examined. The low frequency components based on Hilbert-Huang Transformation are used for identifying finless porpoise.

The passive acoustic system and the recognition method can be used for long-term monitoring finless porpoise in fixed location and surveying finless porpoise in a mobile method. The new automatic recognition method of finless porpoise has great practical significance for surveying and protecting the Yangtze finless porpoise in the wild.

More Yangtze finless porpoise data will be researched in the future. It is very meaningful to understand the acoustic signals emitted from finless porpoises. If the meaning of finless porpoise signal can be understood, the behavior of the finless porpoise will be perceived, and talking to finless porpoises can be possible. That will be much more useful for protecting this endangered species.

Contact Us
  • 86-10-68597521 (day)

    86-10-68597289 (night)

  • 86-10-68511095 (day)

    86-10-68512458 (night)

  • cas_en@cas.cn

  • 52 Sanlihe Rd., Xicheng District,

    Beijing, China (100864)

Copyright © 2002 - Chinese Academy of Sciences