Automated real-time recognition of sounds is a powerful means to inform rangers of potential illegal activities in large nature reserves. Many illegal activities are accompanied with noise, varying from engine sounds to gunshots, and from cattle to dog sounds. They can all be used to alert rangers.
Machine listening is a very new and fast-developing field of expertise, heavenly based on machine learning. To design and deploy a sensor capable of recognising many different types of sounds in realtime, many engineering hours and development cycles are needed. Hence, we chose to develop it in the jungle which we call our home, before we deploy it in the actual jungle.
Together with volunteers of the IoT Sensemakers Amsterdam, we have improved the first prototype of our SERVAL sound sensor. Meanwhile we started a citizens science project with the City of Amsterdam, the Ombudsman, and De Waag, to address the problem of noise pollution.
This Thursday we started sound-sampling the Amsterdam City jungle! Together with friends from de Waag and IoT Sensemakers Amsterdam we collected many hours of recordings of sounds that are known to bother people.
This summer we’ll use the collected samples to train the SERVAL sensor, after which it will be deployed in the city to assess and address noise pollution in a much more meaningful manner then ever before.
The first results are expected by the end of this year. Once the technology has been proven in the city jungle, we’ll bring it to the jungles and savannes of Africa and Asia to help protecting wildlife.
We can’t wait!