An Automatic Human Fall Detection Approach Using RGBD Cameras

Shugang Zhang,Zhen Li,Zhiqiang Wei,Shuang Wang
DOI: https://doi.org/10.1109/iccsnt.2016.8070265
2016-01-01
Abstract:Unexpected fall remains the major threat for elderly since it can cause severe consequences. In this paper, an automatic human fall detection approach is proposed using RGB-Depth (RGBD) cameras. Based on the skeleton and joints data, three velocity features and a head-hip height difference feature are extracted within each sliding window. The features are then gradually analyzed and compared to adaptive thresholds by the proposed multi-step fall detection method. We also conducted a series evaluation on the determination of thresholds and the whole approach. The experiments show a good result with a high F1-measure of 0.944.
What problem does this paper attempt to address?