Integrating Multimodal Sensing and Interpretable AI for Ecological Monitoring of Subterranean Termite Activity

This repository contains the dataset, analysis scripts, and trained model supporting the findings presented in the manuscript "Integrating Multimodal Sensing and Interpretable AI for Ecological Monitoring of Subterranean Termite Activity," submitted to Ecological Informatics ([DOI when assigned]). The study presents a low-cost framework using multimodal sensors (piezoelectric transducers, MEMS microphone, accelerometer) and a lightweight neural network (BANN-Lite) to detect and classify the activity states ('Active' vs. 'Inactive') of subterranean termites (Odontotermes sp.) based on their acoustic signatures, particularly a distinct 0-50 Hz fingerprint.