A researcher at the University of Wollongong has developed a small, low-cost device that can identify dangerous mosquitoes simply by listening to the sound of their wingbeats.
The new technology could help health authorities detect disease-carrying mosquitoes much faster, especially in remote areas where laboratory testing is difficult or unavailable.
The device was created by Dr. Kiran Trivedi and recently showcased at the United Nations AI for Good Global Summit in Geneva.
It uses artificial intelligence (AI) to recognize three of the world’s most important mosquito groups: Aedes, Anopheles, and Culex.
These mosquitoes spread serious illnesses such as dengue fever, malaria, and West Nile virus, which together affect millions of people every year.
According to the World Health Organization, mosquitoes are the deadliest animals on Earth, causing hundreds of thousands of deaths annually. The greatest impact is often seen in developing countries, where access to advanced testing laboratories can be limited.
Traditional mosquito surveillance usually involves collecting mosquito larvae from breeding sites, transporting them to a laboratory, and identifying the species under a microscope. While this method is accurate, it is also time-consuming and requires trained specialists.
Dr. Trivedi’s device offers a much quicker approach. Every mosquito species beats its wings at a slightly different speed, producing a unique sound or acoustic “fingerprint.” By analyzing these wingbeat sounds, the device can identify the mosquito species within seconds.
Unlike many AI systems that rely on powerful computers or internet connections, this device uses a technology called Tiny Machine Learning, or TinyML. This allows the AI model to run directly on a small, energy-efficient computer chip inside the device. As a result, it works without an internet connection, cloud computing, or expensive infrastructure, making it well suited for use in rural and remote communities.
The prototype is built using a small Arduino-based circuit board equipped with a microphone and a display. The AI model was trained using publicly available recordings of mosquito wingbeats and currently identifies mosquito species with an accuracy of about 88%. Dr. Trivedi believes this performance could improve even further as better microphones and higher-quality recordings become available.
He also sees the technology becoming much more powerful when many devices are deployed together. Instead of operating as a single detector, networks of these devices could continuously monitor mosquito populations across entire regions and automatically build live maps showing where disease-carrying mosquitoes are becoming more common.
Dr. Trivedi compares the idea to navigation apps that display traffic conditions in real time. Rather than waiting until disease outbreaks occur, public health officials could spot mosquito hotspots early and take action before infections begin to spread.
If further developed, this simple listening device could become an affordable and practical tool for helping communities around the world detect dangerous mosquitoes sooner, improve disease surveillance, and strengthen efforts to prevent outbreaks before they start.
