Noise pollution is an emerging and challenging problem of all large metropolitan areas, affecting the health of citizens in multiple ways. Therefore, obtaining a detailed and real-time map of noise in cities becomes of the utmost importance for authorities to take preventive measures. Until now, these measurements were limited to occasional sampling made by specialized companies, that mainly focus on major roads. In this paper, we propose an alternative approach to this problem based on crowdsensing. Our proposed architecture empowers participating citizens by allowing them to seamlessly, and based on their context, sample the noise in their surrounding environment. This allows us to provide a global and detailed view of noise levels around the city, including places traditionally not monitored due to poor accessibility, even while using their vehicles. In the paper, we detail how the different relevant issues in our architecture, i.e., smartphone calibration, measurement adequacy, server design, and client⁻server interaction, were solved, and we have validated them in real scenarios to illustrate the potential of the solution achieved.
Keywords: machine learning; mobile crowdsensing; noise-sensing; smart cities; smartphone; weka.