Automatic acoustic identification of insects inspired by the speaker recognition paradigm

This work reports our research efforts towards developing efficient equipment for the automatic acoustic recognition of insects. In particular, we discuss the characteristics of the acoustic patterns of a target insect family, namely the cricket family. To address the recognition problem we apply a feature extraction methodology that has been inspired by well documented tactics of speech processing, which were adapted here to the specifics of the sound production mechanism of insects, in combination with state-of-the-art speaker identification technology. We apply this approach to a large and well documented database of families and subfamilies of cricket sounds, and we report results that exceed 99% recognition accuracy on the levels of family and subfamily, and 94% on the level of a specific insect out of 105 species. We deem this equipment will be of practical benefit for non-intrusive acoustic environmental monitoring applications as it is directly expandable to other insect species.