Signal classification by similarity and feature extraction allows an important application in insect recognition

Insects have a strong relationship with the humanity, in both positive and negative ways. It is estimated that insects, particularly bees, pollinate at least twothirds of all food consumed in the world. In contrast, mosquito borne diseases kill millions of people every year. Due to such a complex relationship, insect control attempts must be carefully planned. Otherwise, there is the risk of eliminating beneficial species, such as the recent threat of bee extinction. We are developing a novel sensor as a tool to control disease vectors and agricultural pests. This sensor captures insect flight information using laser light and classify the insects according to their species. Therefore, the sensor will provide real-time population estimates of species. Such information is the key to enable effective alarming systems for outbreaks, the intelligent use of insect control techniques, such as insecticides, and will be the heart of the next generation of insect traps that will capture only species of interest. In this paper, we demonstrate how we overtook the most importante challenge to make this sensor practical: the creation of accurate classification systems. The sensor generates a very brief signal as result of the instant that the insect crosses the laser. Such events last for tenths of a second and have a very simple structure, consequence of the wings movements. Nevertheless, we managed to successfully identify relevant features using speech and audio analysis techniques. Even with the described challenges, we show that we can achieve an accuracy of 98% in the task of disease vector mosquitoes identification