Estimation of sleeping posture for M-Health by a wearable tri-axis accelerometer

Abstract

Sensor technology has been developed for measuring daily activity. Measurement instruments of all kinds have continued to become much smaller, consume less power, and increase in resolution and/or sensitivity. This study proposed a way to detect sleeping postures from data acquired using a tri-axis accelerometer strapped to subject's chest. We defined sleeping postures as five states focusing on basal-surface, which indicates the place of human's body on the ground. An evaluation was performed both manually by a technician and automatically by designed software. Results is that the accelerometer allowed us to accurately classify postures. Changes in posture were detected with a mean error of less than 3 s, which is acceptable for clinical cardiovascular applications.