Adaptive Notifications Generation for Smartphone Users Based on their Physical Activities

Md. Mobasshir Arshed Naved, Md. Yusuf Sarwar Uddin

Abstract: Doing physical exercises routinely is important for certain health conditions, such as controlling diabetes or hypertension and for rehabilitation. As smartphones are possessed and carried by individuals, their sensors data, for instance data from accelerometers, can be utilized to determine the level of physical activities an individual is undertaking daily and then to generate appropriate suggestions if the individual falls short in achieving the necessary goals. In this paper, we propose an adaptive notification generation system that identifies basic physical activities of a user from his/her smartphone's 3D accelerometer data and then suggests the user through mobile phone notifications the recommended level of physical activities he/she should undergo. We train an activity recognition model from an existing dataset of accelerometers data and then use the trained model to track a user's current level of physical activities. The system automatically generates activity suggestions in case the level of physical activities falls below a certain desired rate and sends the suggestions as mobile phone notifications for user's attention. Our activity suggestion system thus enables a cost-effective behavioral intervention to promote physical activities.