Malaria and Dengue Fever, spread by mosquitos, are a big public health issue in developing countries. If we could detect mosquitos and classify their species, governments and NGOs would be able to more effectively target eradication and treatment resources. Additionally, the technology could be incorporated into smoke alarm style appliances for the home that released a puff of insecticide when an insect was detected, or alerted the householder to the danger.
Preliminary work by Tom Nickson (ORCHID DPhil student) has shown that it might be possible to detect specific species of mosquito by their sound. This work formed part of a joint application with Kew Gardens to a Google backed competition, the Google Impact Fund. The application was successful and the team received £500,000 along with the promise of mentoring and support from Google and the potential to get Googlers involved with the project.
The plan is to fuse Kew’s environmental and biological expertise with Oxford’s abilities in machine learning and signal analysis. From the initial work, it seems unlikely that it will be possible to pinpoint the exact species of mosquito that is present – due to variations between individuals it is not always possible to draw a clear line between species based on the sound alone. A distribution over species from the acoustic data will be fused with prior information about preferred habitats and official species surveys to generate a report on the probable presence or absence of a given species of mosquito. A less accurate classifier that simply distinguishes between “dangerous” and “not dangerous” groups, or even “mosquito present” and “no mosquito” is also an aim and will be more accurate and simpler to implement.
The algorithm will be designed to run on low power embedded micro-controllers and cheap smartphones and will exploit prior research by the ORCHID Cicada Hunt project to make this low power, rugged hardware and efficient software. The generated data will feed into a GIS system from Kew to be fused with environmental factors and presented to decision makers and other interested parties.