Fall Detection Algorithm for Elderly People Living Alone Based on ARKit and YOLOv5
Hui Jin, Yuh-Chung Lin, Odai Athamneh, Weijun Zhou, Huiting Xu · 2024
The article presents a fall detection algorithm for elderly people living alone based on ARKit and YOLOv5. By combining ARKit’s spatial perception capabilities with YOLOv5’s high-speed visual analysis, the algorithm can quickly and accurately identify elderly people falling. It comprises functional modules such as data collection and fall detection. It adopts the operating environment of the iOS platform and is developed using Swift, Object-C++, and Python languages. The algorithm demonstrates strong fall detection performance with high accuracy, recall, and mAP. However, limitations in the versions of YOLO as well as the size and diversity of the dataset. Future work could focus on using different YOLO versions, expanding the dataset, changing the network structure, and introducing more techniques to improve the accuracy and performance of the algorithm.