Automated identification of mosquito (Diptera: Culicidae) wingbeat waveform by artificial neural network

Abstract

The wingbeat waveforms of Aedes albopictus (Skuse), A. aegypti (Linnaeus), Culex pipiens pallens (Coquillett), C. pipiens quinquefasciatus Say, and C. pipiens molestus Forskal were recorded by a photosensor and WfRer system. Wingbeat frequencies were extracted from the wingbeat waveform. Back Propagation artificial neural network classifiers were built to identify automatically the species of mosquitoes by wingbeat waveform, wingbeat frequencies, and combining them as input variables. The most accurate classifier tested was an artificial neural network by wingbeat frequencies as input variable. The accuracy was average 72.67% and highest 89.00%. It also demonstrated that it was possible to identify automatically the species of mosquitoes. Meanwhile, it probably depended on the characteristic of frequencies for identifying each other among different mosquitoes. © 2005 by International Federation for Information Processing.