Research on Fall Detection Method of Empty-nesters Based on Computer Vision
Yaochang Xi, Peijiang Chen, Zhenxing Fu · 2022
Due to the frequent occurrence of falls of empty nesters, a timely and accurate fall detection method is designed based on computer vision. Using background subtraction, the moving objects are detected from videos, and then the thresholds of four fall features of the human body are set, including width-height ratio, centroid change rate, effective area ratio, and inclination angle. If the threshold condition is met, the body fall is determined. Combined with multi-characteristics, the method accurately identifies non-fall behaviors such as walking, squatting, and sitting down. The experimental results show that the fall detection algorithm based on multi-feature fusion accurately judges the fall. The algorithm has the advantages of a low amount of computation and satisfiable robustness.