Skeleton-based Fall Events Classification with Data Fusion

Human fall detection aims to classify falls and normal activities. It can improve the speed of rescue for the elderly after a fall occurs, it can also efficiently prevent the elderly from suffering secondary injuries due to untimely or inaccurate fall detection. This technology is widely used in hospitals, smart homes and nursing homes. The challenges are that the injured parts of the elderly can vary due to different types of fall events, such as falling sideways and falling backwards. In this paper, we propose a data fusion method which combine the skeleton keypoints captured from RGB images into fused keypoints to improve the performance of fall events classification. Four well known classification methods, Random Forest, Support Vector Machine, Multi-layer Perceptron, AdaBoost are used in the proposed framework. Meanwhile, the impact on fall detection results due to missing data caused by occlusion and or privacy protection is also analyzed through ablation study. The experimental results confirm that the proposed framework outperformed the state-of-the-art with reduced computational cost.