A Survey of Sound-based Biometrics used in Species Recognition

Preserving animals and, in particular, endangered animals is of paramount importance which is pointed out by environmentalists, biologists and other life scientists frequently. In tracking and monitoring such animals, people often use visual clues such as shape, color and physical attributes to search for species of interest. Such clues while useful, may prove ineffective at times due to weather condition, pollution and other natural phenomena that may impair the acquisition of a clear image/video. This paper focuses on developing audio signatures based on sounds made by animals of interest and utilizing these signatures to locate and track such animals. The problem addressed here may be viewed as a behavioral biometrics problem as the goal here is to automate classification/identification of species based on their audio signatures. Here, the theoretical basis for developing a species recognition system using the sound produced by such species is addressed. Acquisition of sound data, preprocessing, filtering and finally classification of such data are described in detail.