Performance Evaluation of a System to Detect Awakening Behavior of Bedridden Patients with a Quilt Using a Depth Sensor

We evaluate a system that monitors bedridden patients covered by different types of quilts by utilizing a proximity depth sensor with an infrared projector (Kinect). This system is capable of detecting people as they wake up. We believe that this system could be useful for monitoring patients in hospitals or private homes who must occasionally be left unsupervised. The system, which utilizes an optimal linear discriminant function, is demonstrated with subjects covered by quilts of different thicknesses. The detection rate of a basic system using height and space data for four cases - lying on the back, lying on the side, sitting up, and no longer in bed—was 99.2 % when a thin quilt was used and 80.4 % when a thick quilt was used. The detection rate was also low (80.8 %) when two variables were used for the cases using either thin or thick quilts. The rate improved to 96.3 % using the data of the initial condition of the subject. Good performance was obtained regardless of the thickness of the quilt