University of Wollongong TinyML Device Identifies Disease‑Carrying Mosquitoes

Researchers at the University of Wollongong, led by Associate Professor Kiran Trivedi, have created a low‑power TinyML device that can differentiate mosquito species by analysing the unique pitch of their wing‑beat sounds. By converting 40 GB of audio recordings into spectrogram images, the AI model learns to recognise the low‑pitch buzz of Anopheles (malaria), the medium‑pitch of Culex (Japanese encephalitis, Ross River virus) and the high‑pitch of Aedes aegypti (dengue, Zika). The model, originally 150 MB, was compressed to fit a tiny chip suitable for field deployment.

The technology builds on earlier TinyML prototypes that identified bird calls and interpreted infant cries, but the mosquito application targets the roughly 700,000 annual deaths caused by vector‑borne diseases. The prototype was showcased at the UN AI for Good summit, highlighting its potential to support public‑health surveillance in endemic regions.