sEMG-Based Motion Recognition of Upper Limb Rehabilitation Using the Improved Yolo-v4 Algorithm
Dongdong Bu1, Shuxiang Guo1,2, He Li1
1Key Laboratory of Convergence Biomedical Engineering System and Healthcare Technology, The Ministry of Industry and Information Technology, School of Life Science, Beijing Institute of Technology, No. 5, Zhongguancun South Street, Haidian District, Beijing 100081, China.
Life (Basel, Switzerland)
|January 21, 2022
Summary
This study introduces a new method for controlling upper limb exoskeleton robots using surface electromyography (sEMG) images. The approach enables faster, more accurate motion recognition and joint angle prediction for improved rehabilitation.
Area of Science:
- Biomedical Engineering
- Robotics
- Signal Processing
Background:
- Surface electromyography (sEMG) is crucial for controlling upper limb exoskeleton robots.
- Traditional sEMG control methods involve complex feature extraction, hindering real-time application due to computational demands and individual differences.
Purpose of the Study:
- To develop a novel, high-security, and fast-processing action recognition strategy for exoskeleton robot control.
- To enable real-time limb joint motion recognition and joint angle detection using sEMG images, overcoming limitations of traditional methods.
Main Methods:
- Utilized MobileNetV2 with the Ghost module for feature extraction, creating a pre-trained model.
- Employed the Yolo-V4 target detection network to classify six upper limb joint movement categories and predict joint angles from sEMG images.
Main Results:
- The proposed method achieved an average recognition accuracy of 80.7% on verification data.
- Processing speed reached 17.97 ms per image on a PC, with 78 out of 100 images accurately identified.
- Achieved mAP@0.5 of 82.3% and mAP@0.5-0.95 of 0.42 on training data.
Conclusions:
- The developed sEMG image-based motion recognition strategy is effective for controlling upper limb exoskeleton robots.
- This novel approach offers a viable solution for real-time, accurate control, enhancing exoskeleton robot applications in rehabilitation.
