Towards an Automated Acoustic Detection Algorithm for Wood-Boring Beetle Larvae (Coleoptera: Cerambycidae and Buprestidae).

The development of acoustic systems for detection of wood-boring larvae requires knowledge of the features of signals produced both by insects and background noise. This paper presents analysis of acoustic/vibrational signals recorded in tests using tree bolts infested with Anoplophora glabripennis (Motschulsky) (Coleoptera: Cerambycidae) (Asian longhorn beetle) and Agrilus planipennis Fairmaire (Coleoptera: Buprestidae) (emerald ash borer) larvae. Based on features found, an algorithm for automated insect signal detection was developed. The algorithm automatically detects pulses with parameters typical for the larva-induced signals and rejects noninsect signals caused by ambient noise. The decision that a wood sample is infested is made when the mean rate of detected insect pulses per minute exceeds a predefined threshold. The proposed automatic detection algorithm demonstrated the following performance: 12 out of 15 intact samples were correctly classified as intact, 23 out of 25 infested samples were correctly classified as infested, and five samples out of the total 40 were classified as 'unknown.' This means that a successful wood-sample classification of 87.5% was achieved, with the remaining 12.5% classified as 'unknown,' requiring a repeat of the test in a less noisy environment, or manual inspection.