There are tens of thousands of plants, animals and fungi in the Netherlands alone. To learn how to distinguish them can seem like a mountain climb. Apps that can use Artificial Intelligence to identify species can greatly lower this threshold. For instance, for photos of Dutch and European species, there is ObsIdenfity. For bird sounds, there is Birdnet and Merlin. However, birds are not the only animal species that make sounds: many of the Dutch grasshoppers and crickets produce sounds, which vary by species. Crickets make whistling sounds, grasshoppers usually make more scratching, tapping or rattling sounds.<\/p>\n
To make learning these sounds much more accessible, Aquila Ecology, in collaboration with Province of Limburg, has set about creating an Android app that uses AI to distinguish the sounds. The 35 species the app can recognise are distinguished with 90% accuracy, which is exceptionally high. The app, when it becomes available, will show pictures of the species upon recognition and a text with basic info.<\/p>\n
With the accurate recognition of sounds, it suddenly also becomes possible to analyse long recordings and recognise all species in them. This allows all kinds of studies on the occurrence and activity of grasshoppers that were not possible before.<\/p>\n
That the recognition is so accurate is due to an algorithm developed by Aquila Ecology specifically for this purpose. Whereas birds are recognised on the basis of spectrograms (visualisation of pitches over time), Aquila Ecology's algorithm works on the basis of raw sound data. Developing such an algorithm is a lot of work: there are endless possibilities of different layers of operations that eventually lead to a recognition. But the challenge is also to stack the right layers on top of each other to get the best possible results.<\/p>\n