Research project between SINTEF and Norsonic – Construction sites reported as being too noisy will be relieved

Acoustics experts led by Femke B. Gelderblom from the Norwegian research centre SINTEF https://www.sintef.no have developed “NoiseTag”, an AI-based microphone. “AI analysis of sound is still quite underdeveloped compared to what AI can achieve with images or text. Recognising that a particular sound comes from the same source as a similar sound originating at a greater distance can quickly overwhelm an AI model,” says the expert.

Reliably detecting external noise

The scientists have been working together on a research project with Norsonic https://www.norsonic.com. The company, also based in Norway, manufactures devices for measuring noise and vibrations. The result is a completely new device. Until now, noises that distort a measurement, such as motorbike noise, have had to be eliminated from the measurement by humans listening to the recordings.

NoiseTag now takes care of this. “We have trained the system to eliminate extraneous noise independently,” says Norsonic software engineer Karl Henrik Ejdfors. He presented the AI with potential extraneous noises such as motorbike noise or seagull cries and instructed the AI not to take these into account during noise recording.

Noise AI is getting better and better

“In noise monitoring, it is crucial to distinguish between what is relevant noise and what is not. A construction project is very often located next to a main road. The road is noisy, but that does not, of course, mean that the construction machinery is making too much noise. In future, we will be able to take this into account without having to listen to the recordings,” says Gelderblom. The device has just been launched and is set to improve continuously through machine learning.