MURAGAKI Yoshihiro    Department   School of Medicine(Tokyo Women's Medical University Hospital), School of Medicine    Position   Visiting Professor
Article types	Original article
Language	Japanese
Peer review	Peer reviewed
Title	Bed Exit Detection Using Depth Image Sensor
Volume, Issue, Page	pp.45-53
Publication date	2014/04
Summary	Falls are a frequent cause of unintentional serious injury to patients, and a number of studies on this important issue have been conducted. Preventing inpatients from falling is very difficult, as it poses a heavy burden on the medical staff. Although sensors can be used to warn of patient falls, they have many false detections, which also burden the medical staff. In this paper, we propose a bed-exit alarm using a depth image sensor. A depth image sensor generates depth images by detecting infrared patterns in images captured using a camera. Using this sensor, patient movement can be captured without attaching any sensors to the patient. The proposed method works as follows. The input image is converted to a three-dimensional point cloud. Bed position and direction are then estimated from the point cloud using the iterative closest point algorithm. Patient movement using a motion vector search is also estimated. Using these results as features, a detection parameter is optimized using a support vector machine. The proposed method was able to detect a patient's bed exit 63 out of 68 times in experimental images. There were 24 false detections. With current sensors having a false detection rate of 70%, our proposed sensor more accurately detects bed exit.
NAID	110009770438