Post by Delft University of Technology

Noise pollution can seriously impact our health. Research links it to stress, sleep problems, high blood pressure, even heart attacks. ‘All the more reason to map what our cities really sound like’, says researcher Lion Cassens. Current noise maps, based on simulations, are not precise enough at city level and therefore give a distorted picture. They measure mainly traffic noise and only at limited locations, not in lively city centres where people walk, talk, and live. They focus too much on decibel levels, not on how people experience sound. ‘Models calculate averages, but that doesn’t reflect reality. That is what I want to change with my research.’ Cassens together with an electrical engineering student built a sensor of about €50 each to collect high quality data, creating a more complete picture of the soundscape. Forty households in Delft installed these low‑cost, solar-powered sensors that recorded sound levels eight times per second. Using AI, the sensors could distinguish 12 sound types, from birdsong and church bells to crowds and traffic, without ever storing raw audio. Thanks to privacy‑by‑design, conversations were never recorded. The survey amongst households indicated that talking and construction noise were more annoying than traffic. This shows the value of AI powered sensors: A traditional measuring station can tell us how loud a sound is but keeps us guessing what causes the noise. Cassens further suspects that short, irregular sound bursts, like that of a loud scooter, are more annoying than constant humming (like highway noise). While the sensor can detect these differences, the survey sample was too small to draw firm conclusions, so more research into this is needed. Cassens work shows that low‑cost AI-tools and resident involvement result in a more accurate soundscape and better insight into impact on residents. “Ultimately I hope my research contributes to a more complete picture of the soundscape to create healthier, more liveable cities.” #university #technology #innovation #engineering #science #research