Identification of Insects Using an Audio Analysis-Based Machine Learning Technique

Abstract

The longstanding interaction between insects and humans has been valuable and lethal. Each year, insects transmit diseases that inflict property damage, but they also fertilize a significant portion of the global food supply. Similarly, new technology is necessary to reduce the number of harmful insects without damaging populations of beneficial insects. In the scope of this study, we offer an intelligent insect catcher that uses a sensor and sound analysis to capture and classify insects. We utilize audio research and machine learning technologies for the insect identification competition to assess many feature sets and create individualized training programs for each participant. We evaluated our classifiers in linear discriminant and binary class settings and discovered that a class labels classifier detecting mosquito species has near-perfect accuracy. It proves that the proposed trap is effective.