A team of researchers have developed an AI system that could be used as a mosquito early warning system that generates an alert when the insects are close, by detecting the sound of their wings. The system can be baked into a phone app and can analyse sound waves and issue a warning when it hears the buzz of a mosquito.
Currently the app can detect when the insects are close, however next in their roadmap is to be able to distinguish between the sound of regular mosquito and a malaria-carrying mosquito.
Yunpeng Li, a machine learning student at Oxford University who worked on the project, said “If we can identify the species, we can tell people in areas where there is malaria that these mosquitoes are around and that they need to take care, to use bed nets and so on.”
The Oxford University team built the AI tool by recording the sound of mosquitoes in a lab, they also gathered more audio signatures from the US Centers for Disease Control and Prevention, scientists working in the forests of Thailand and an army research unit in Kenya. Once they had the dataset of recordings, the team converted the audio signals into frequency features and trained an ML algorithm to learn the pattern created by mosquitoes in flight.
Tests using inexpensive smart phones found that the app could detect the presence of mosquitoes from 10cm, this varied depending on background sound. The system is most effective when the phone is placed somewhere that would attract mosquitoes such as a lightbulb.
Although the sound of mosquito does change as it gets larger, Li is confident that the algorithm can learn the signature produced by different species and sizes. Features such as the wing size and shape feed into the overall sound the mosquito makes.
In order to improve the AI, the team of researchers are now planning on recording 100’s of hours of recordings in the lab and in the wild. They hope to crowdsource the training of the algorithms by asking humans to listen to two-second audio clips and note whether they can hear a buzzing sound.