Home sleep monitoring based on wrist movement data processing

Abstract In this paper, two original sleep monitoring algorithms, including threshold and k-means clustering algorithms are presented. All the proposed algorithms use only acceleration data acquired from the non-dominant wrist with a 3-axis accelerometer, allowing the detection of falling asleep and waking up and a classification into 4-sleep stages (“awake”, “light sleep”, “deep sleep” and “REM”). We validate the proposed methods by comparing them to the results of “Fitbit Charge 2” and “Withings Sleep Analyzer”. Based on wrist movement data collected during 10 nights of sleep of a volunteer, we can show that the algorithms obtain promising results that allow us to consider a new non-intrusive method for users and medical staff to follow the trend of sleep quality through long term monitoring. This longitudinal monitoring can help to detect abnormal changes in sleep that are usually a sign of a change in health status.