Inferring the Urban Noise Pollution with Sparse Data through Crowdsensing

Noise Pollution is a serious problem in modern cities. In recent years, there have been multiple initiatives funded by the European Union (EU) to monitor noise pollution by gathering data about urban sound levels. Recent projects adhere to the Mobile Crowdsensing (MCS) paradigm and use microphones embedded into their mobile phones to record the amount of noise. However, data sparsity in time and space can hinder the effectiveness of the monitoring process. In this paper, we propose an MCS system for noise pollution monitoring that can predict the level of noise in road segments even when data is missing, by leveraging either historical records or spatial correlation between different streets. We evaluate our prediction model against two datasets: one collected by us through a dedicated mobile app, and the other obtained from the literature, showing the effectiveness of our proposal.