Device-free sleep stage recognition through bed frame vibration sensing

Abstract

Sleep disorder impairs people's health. To better analyze user's sleep quality, it is important to monitor people's sleep stages. Prior methods, such as polysomnography (PSG) and Fitbit, are often intrusive and having potential to change user's daily sleep routine. We present a non-intrusive approach to identify sleep stages through bed-frame vibrations. The proposed system detects the movement of users during their sleep and estimates their sleep stages via their movements induced vibration on the bed frame. Our system extracts features from the vibration signal and distinguishes sleep stages between rapid eye movement (REM) and non-rapid eye movement sleep (NREM). We evaluated our system in two ways 1) to understand the data distribution over different sleep stages, and 2) to explore the feasibility of distinguishing different movement characteristics over different sleep stages. The system achieved an area under curve (AUC) score of 0.84 when classifying REM sleep stages over NREM stages.