Intelligent diagnosis of gait disorders using video-based 3D motion analysis

Abstract

This experiment proposed an intelligent disease diagnosis method based on gait video analysis. Gait videos were analyzed using OpenPose for 2D pose estima tion, and a Temporal Convolutional Network (TCN) was employed to predict 3D poses, obtaining 3D gait motion data of the target. After extracting motion features, a knowledge base and inference rules for gait abnormalities and diseases were constructed within the framework of the Open Semantic Technology Intelligent System (OSTIS). The gait features were semantically processed accordingly. Finally, a classification model was used to diagnose potential diseases and provide interpretable diagnostic recommendations. Experimental results demonstrated that this method effectively integrates 3D motion features with semantic reasoning, achieving accurate disease classification and diagnosis, thus offering a novel technological approach to intelligent medical diagnosis.