Vision-Based Gait Analysis Method in Parkinson's Disease

作者：Ziyu Liu, Chengzhang He, Fei Hu Zhang, Junfeng Ge · 年份：2024 · DOI：10.1109/cei63587.2024.10871802 · 被引用次数：4 · 研究领域：Hand Gesture Recognition Systems、Parkinson's Disease Mechanisms and Treatments、Muscle activation and electromyography studies

This paper presents a gait analysis method in Parkinson's disease (PD) based on the Azure Kinect DK camera and human pose estimation algorithms, specifically RTMPose and MotionBert. The method extracts 3D human key points from RGB-D video to analyze gait parameters such as step length, step time, gait speed, arm swing magnitude, gait cycle, and turning duration. The results demonstrate significant differences in these gait parameters between Parkinson's disease patients and healthy individuals, and the method can analyze the patient's abnormal gait. The method offers a non-invasive, real-time tool for PD symptom assessment, with potential for low-cost monitoring and clinical application.