An Improved YOLOv5 Algorithm for Elderly Fall Detection

Abstract

Accidental falls are one of the major threats to the elderly population. Older adults who are not caught in time after a fall may miss the best time to be rescued. We propose an improved YOLOv5 fall detection algorithm in this paper. Firstly, we use SIoU Loss instead of CIoU Loss to improve the speed of training and the accuracy of inference. Secondly, the SimAM attention mechanism is introduced into the YOLOv5 model to further improve the model's performance. Experiments show that the [email protected] of the improved YOLOv5 fall detection model reaches 96.7%, which is 2.4% higher than the original YOLOv5. The overall detection effect is better than the original YOLOv5 model and other network algorithms. The algorithm used in this paper can detect fall behavior promptly, which largely alleviates the risk of death or long-term treatment.