RGB-D camera-based activity analysis

In this paper, we propose a new activity analysis framework to facilitate the independence of elderly adults living in the community, reduce risks, and enhance the quality of life at home by using RGB-D cameras. Our contributions include two aspects: 1) recognizing 5 activities related to falling including standing, fall from standing, fall from sitting, sit on chair, and sit on floor. The main analysis is based on the depth information due to the advantages of handling illumination changes and identity protection. If the monitored person is out of the range of 3D camera, RGB-based video analysis module is employed to continue the activity monitoring. 2) Identifying the monitored person if there are multiple people in the camera view by combining both depth and RGB information. We have collected a dataset under different lighting conditions and ranges. Experimental results demonstrate the effectiveness of the proposal framework.