Awakening Detection System for Patients Covered by a Quilt Using a Depth Sensor and a Neural Network

Abstract

An awakening detection system for bedridden patients covered by a quilt using a ceiling-mounted depth sensor is evaluated. The performance of the depth sensor for monitoring situations where subjects are covered by quilts of different thicknesses using a neural network is discussed. The state of the subject was differentiated into four cases: “lying on back in bed,” “lying on side in bed,” “sitting up in bed,” and “no longer in bed.” The network obtained from the dataset of a specific quilt cannot be used for discrimination in a general case with quilts of different thicknesses. The network obtained using datasets with quilts of different thicknesses shows a good performance. Further improvements in the performance can be obtained by including the values of the maximum height and thickness of the quilt in the datasets