The calls of many animals have evolved to be species specific which has been demonstrated by lots of researches before. Call sonogram statistic analysis and playback experiments in many animal taxa have even shown that the calls of the individuals were individual specific in the same species. Thus researchers can identify animal species and individuals using the acoustic features of animal calls. Researches in this field have been in the rise, but most experiments used long-term statistical features of a specific syllable in the calls, which called call-dependent recognition, and focused on species identification rather than individual recognition. The call-dependent recognition is time-consuming and can not be applied to automatic system easily. In this experiment the researchers did call-independent bird individual identification in four passerines using the most popular acoustic feature Mel-frequency cepstral coefficients borrowed from speech and speaker recognition and a powerful pattern recognition method Gaussian mixture models which is the first use in individual recognition. And the recognition accuracy was 89.1-92.5%. In this paper the researchers constructed a call-independent and automatic acoustic system for the individual recognition of animals which can be easily applied to other species.

This work was completed cooperatively by the researchers of the Biodiversity Informatics Group and Avian Ecology Group, Key Laboratory of Animal Ecology and Conservation Biology and has got supports of the Ornithological Research Group, Key Laboratory of Zoological Systematics and Evolution, Institute of Zoology, CAS. The results will be published in Pattern Recognition.

Jinkui Cheng, Yuehua Sun, Liqiang Ji, 2010. “A call-independent and automatic acoustic system for the individual recognition of animals: A novel model using four passerines” Pattern Recognition 43 (11): 3846-3852

Internet access: http://dx.doi.org/10.1016/j.patcog.2010.04.026