AI Noise Sensor

How did you come up with the idea for AI Noise Sensor?

A noisy neighbour and motorbikes without silencers gave me inspiration to investigate ‘noise recording’. I wanted an automated solution that would monitor noise pollution round-the-clock, and also identify sources of noises – whether it was loud music, motor vehicles, construction work etc.

How did you go about developing the solution?

I trained a Convolutional Neural Network (CNN) with 8000+ noise audio clips, which I had converted to image spectrograms. I saved the CNN model into a Raspberry Pi (configured in headless mode with a mic), and wrote programs to record noises in real time, measure loudness, and predict noise source.

How does AI Noise Sensor work?

The Raspberry Pi when plugged in starts 5 of my programs that records audio and plots all information to a graphical interface, looking at which Environment Managers identify dominant sources of noises, the times at which they occur, the durations for which they occur, and their loudness

Why should your product win its category and why should people vote for you in the People’s Choice Award?

Noise pollution is a silent (ironically) threat that causes mental and physical health problems. My AI product will help the general public and environmental managers identify sources of noise, and hence tackle the problem at its origin. My product is low cost (£25), portable, and plug n play.