A team from the University of Wollongong (UOW) has built an inexpensive device that is capable of identifying disease-carrying mosquitoes based on the sound of their wingbeats.

The device — which promises to offer a rapid alternative to existing surveillance methods used for tracking malaria and dengue — relies on artificial intelligence (AI) to identify mosquitoes belonging to three of the most significant disease-carrying groups: Aedes, Anopheles and Culex.

Source: University of Wollongong

According to the device’s developers, this identification can occur in mere seconds and without an internet connection, making it ideal for developing nations and remote communities where lab resources are limited, or nonexistent, but necessary for identification.

Existing mosquito surveillance requires collecting water samples from breeding sites and analyzing larvae in the lab to identify the species. To expedite identification, the researchers suggested that sound could perform the same function but faster because each mosquito species beats its wings differently, thereby producing a subtle but distinct acoustic fingerprint.

To accomplish this, the device used Tiny Machine Learning (TinyML), which enabled AI models to run directly on a compact, low-power chip instead of relying on cloud computing. This allowed the system to identify mosquitoes within seconds while operating offline, thus reducing costs and eliminating privacy concerns.

The model, which was trained on publicly available recordings, achieved a roughly 88.3% rate of accuracy during trials. That percentage is expected to increase with better microphones and cleaner recordings, the team noted.

"Just as a navigation app shows you traffic in real time, this could show where disease-carrying mosquitoes are building up," the researchers concluded. "Instead of waiting for an outbreak, communities and public health agencies could see the hotspots early and respond."