Urban noise mapping with a crowd sensing system

Abstract

Noise pollution poses a serious threat to people living in cities today. To alleviate the negative impact of noise pollution, an urban noise mapping can be helpful. In this paper, we present the design of NoiseSense, a crowd sensing system for housing a real-time urban noise mapping service. A major challenge in building such a system is caused by the sparsity problem of the limited noise measurement data from smartphones. To tackle this challenge, we propose a hybrid approach including a neighborhood-based noise level estimation method and a semi-supervised tensor completion algorithm for inferring noise levels for locations without measurements by smartphone users. This approach leverages a variety of urban data sources, such as Point of Interests, road networks, and check-in data. We also provide a noise prediction method for forecasting the noise levels in the next few hours. We implemented the system and developed an APP for smartphone users. We conducted experiments and field study. The experimental results show that the proposed approach is superior in inferring noise levels merely with sparse measurements from smartphone users. And the prediction approach also outperforms other baseline methods.