Automated Depth Video Monitoring for Fall Reduction: A Case Study

Josh Brown Kramer, Lucas Sabalka, Ben Rush, Katherine Jones, Tegan Nolte; Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, 2020, pp. 294-295

Patient falls are a common, costly, and serious safety problem in hospitals and health care facilities. We have created a system that reduces falls by using computer vision to monitor fall risk patients and alert staff of unsafe behavior before a fall happens. This paper is a companion and followup to "Modeling bed exit likelihood in a camera-based automated video monitoring application," in which we describe the Ocuvera system. Here additional details are provided on that system and its processes. We report clinical results, detail practices used to iterate rapidly and effectively on a massive video database, discuss details of our people tracking algorithms, and discuss the engineering effort required to support the new Azure Kinect depth camera.