Lower Limb Rehabilitation Motion Angle Measurement Based On Deep Learning YOLOv3 Model
Konferenz: ICMLCA 2021 - 2nd International Conference on Machine Learning and Computer Application
17.12.2021 - 19.12.2021 in Shenyang, China
Tagungsband: ICMLCA 2021
Seiten: 6Sprache: EnglischTyp: PDF
Persönliche VDE-Mitglieder erhalten auf diesen Artikel 10% Rabatt
Autoren:
Yang, Yunlong; Yin, Hesheng (State Key Laboratory of Robotics and System, Department of Mechanical Engineering, Harbin Institute of Technology, Harbin, China)
Wang, Liancheng (WEIHAI FURUI ROBOTICS Co., LTD, Weihai, China)
Yao, Yufeng (Department of Mechanical Engineering, Harbin Institute of Technology (Weihai), Tianzhi Institute of Innovation and Technology, Weihai Economic and Technological Development Zone, Weihai, China)
He, Chenxi (Department of Electrical Engineering and Automation, Harbin Institute of Technology (Weihai), Weihai, China)
Inhalt:
The aging of the population and the high incidence of hemiplegia have led to an increasing demand for easy-to-use rehabilitation training. The feedback sensing system which can measure and analyze the lower limb rehabilitation motions is highly significant for improving the rehabilitation outcome. Computer vision-based human motion angle measurement has attracted significant interest. This study aims to measure and analyze the lower limb motion angle in the sagittal plane with a single RGB camera. This paper proposes a method for extracting and monitoring of the lower limb marker points based on YOLOv3 and DarkNet-53 convolutional neural networks, and optimizes the pixel coordinates of the target point based on Kalman. The measurement accuracy of the proposed method is tested by JACO robotic arm, and the test shows that the standard deviation (SD) of the measurement is less than 0.5deg.