Monitoring Sleep and Detecting Irregular Nights through Unconstrained Smartphone Sensing

Sleep is essential for a person's health and well-being. Recent advances of wearable devices and smartphone sensing have led to the proliferation of at-home sleep monitoring solutions for the consumer market. In this paper, we study how to monitor basic sleep behavior and how to detect irregular sleep nights, through unconstrained smartphone sensing, which can serve as an important indicator for both mental and physical health if the sleep problems persist. We first propose a supervised learning approach to predict bedtime and sleep duration with a light-weight context sensing schedule to minimize battery consumption. The proposed solution is validated through an extensive user study, and the prediction accuracy of bedtime and sleep duration significantly outperformed the state-of-art solution. In addition, we propose an unsupervised approach to detect irregular sleep nights by profiling and detecting contextual variations. The experiment results show that the proposed solution is effective in detecting irregular sleep nights. To the best of our knowledge, this is the first work that uses unconstrained smartphone sensing to detect sleep pattern changes with the benefits of reduced training efforts and improved robustness against behavior diversity.