Fall Risk Pose Pattern Analysis Using GNN-based Association Rule Mining

This study proposes a GNN-based association rule mining method for predicting fall risks. To overcome the limitations of wearable devices, pose data was collected using computer vision techniques and integrated into a GNN model for precise predictions. The use of DeepPose for pose estimation, coupled with FP-growth for association rule extraction, followed by the application of a GAT model, enables the learning of complex interactions between poses. The proposed method demonstrates superior performance, achieving over 98% in Accuracy, Precision, Recall, and F1-Score compared to OpenPose-based approaches. This model highlights the potential for enhancing fall prevention and real-time monitoring in high-risk groups.