MOTIA3: Technical report WP4 activities

Work package 4 (WP4) of the MoTIA3 project focused on developing semi-supervised algorithms based on deep learning for the detection of sound anomalies, applied to the analysis of the flight of tiger mosquitoes (Aedes albopictus) in the Sterile Insect Technique (SIT). This technique, used in the Valencian Community for biological pest control, requires a sexing process to release only sterile males. It was proposed to use mosquito flight sound analysis as a quality control, since the flight sound of males and females is different. The SPADE algorithm, originally designed for imaging, was adapted to sound anomaly detection by developing new architectures and optimizing parameters to address domain shift and achieve robust models to variations in the data, such as age or state of the mosquitoes. The results demonstrated the feasibility of the approach to differentiate between males and females, providing a useful tool to improve quality control in SIT. In addition, the proposed solutions have potential applications in other sectors, such as industrial predictive maintenance, contributing to the advancement of sustainable and efficient technologies based on artificial intelligence.