Air Pollution Exposure Estimation and Finding Association with Human Activity using Wearable Sensor Network

Abstract

Air quality and pollution monitoring services are provided by many countries and cities. However, individuals are more concerned about personal exposure and dosage, which can rarely be estimated due to the low spatial resolution of air pollution data and lack of personal data. In recent years, an increasing number of research groups, including ours, have focused on increasing the spatial resolution of air pollution data using ubiquitous sensor networks. These works did raise the spatial granularity compared with data from fixed air pollution monitoring sites. In this paper, we combine air pollution and human energy expenditure data to give individuals real-time personal air pollution exposure estimates. In particular, this paper describes our experiences with developing a personal air pollution exposure estimation system utilising participatory air pollution monitoring system and energy expenditure data collected from wearable activity sensors. Our system and applications will benefit the understanding of the relationship between air pollution exposure and personal health. We also conducted a trial to get a full day's air pollution inhalation dosage for one participant, and applied multiple data mining techniques to find out associations between activity mode, location, and the inhaled pollution. Results show that sleep, having meals, working in a campus, and general home activities like reading books will lead to a low air pollution dosage, while working out, walking and driving will cause higher inhaled dose. Furthermore, classification results in our study based on activity modes, locations and dosage data which is collected in the trial show that up to 94% classification accuracy can be achieved.