Insects Sound Classification with Acoustic Features and k-Nearest Algorithm

Abstract

Monitoring particular species and the health of the entire ecosystem necessities determining the existence and number of insects. Several insects are most easily detected by their sounds, and thus, passive acoustic monitoring is highly appropriate. However, practical constraints including the requirement for a human setting, dependency on example sound libraries, low accuracy, low robustness, and limited ability to generalize to novel acoustic situations frequently prevent the advancement of acoustic monitoring. Here, we report outcomes from a collaborative data challenge. The study has utilized improved acoustic monitoring datasets, summarizes the machine learning methods put forth by challenge teams, and carry out extensive performance analysis. The study includes different machine learning models and the study has found 85.4% accuracy from K nearest neighbor method. In the future, this study will be extended to remote monitoring projects. The study also needs to validate more sound features with the help of modern artificial intelligence models.