Falling Pattern Analysis Based on 3D Euclidean Distance of Human Skeleton Joints
Nuth Otanasap, Poonpong Boonbrahm · 2012
There are many elderly people fall each year that have shown in many research such as Global brief for World Health Day 2012. A fall accident not only produces emotional disturbance for the elderly but causes death and symptoms after fall also. The goals of this study are to analysis of falling pattern in pre-fall and critical fall phases based on 3D Euclidean distance of human skeletal joint positions as early detection of fall and comparing the results. The 5 volunteers were observed by a single Kinect camera for real-time 3D skeleton tracking while they walked and fall down to the floor in the range of camera views. The 30 fps input video image collection obtains a set of vectors of 20 skeletal joint positions that include 50 normal lie-down and 50 fall-down postures. The experimental results show that the distance and its derivative between head and left foot and those between head and right foot are good clues for fall detection. As future work we will consider 3-point angles, in addition to distance, for prediction.