A study of automatic classification of sleeping position by a pressure-sensitive sensor

Abstract

Currently in care facilities, a fall preventive movement sensor is often used to prevent falls of care receivers during the night.. But the current sensors are usually designed to detect the motion of care receivers' getting out of bed. Therefore, there are cases where the care receiver has already fallen from the bed by the time the sensor reacted to the movement. It is a common knowledge that a person frequently changes position while sleeping. In this research, we focus on the frequency of sleeping position changes, and aim to realize a method for precise prediction of care receivers' attempt to get out of bed sufficiently before the actual action. We employed the automatic classification method of sleeping position in the pressure-sensitive sensor, with consideration to privacy of the research subjects, and identified nine types of sleeping position that are common, with 77.1% of accuracy. This result is reported in this paper.