Related Experiment Video
Updated: May 28, 2026
11:06
A Human-machine-interface Integrating Low-cost Sensors with a Neuromuscular Electrical Stimulation System for Post-stroke Balance Rehabilitation
Published on: April 12, 2016
sEMG-Based Motion Intention Recognition for Interactive Upper Limb Nursing Assistance
Zekun Peng1, Yongfei Feng2, Liangda Wu2
1School of Mechanical Engineering, Shanghai Jiao Tong University, Shanghai 200240, China.
Sensors (Basel, Switzerland)
|May 27, 2026
Summary
This study developed a robust framework using surface electromyography (sEMG) for recognizing upper-limb motion intentions. The system achieved high accuracy, showing promise for real-time human-machine interaction in assistive devices.
Area of Science:
- Biomedical Engineering
- Neuroscience
- Human-Machine Interaction
Background:
- Surface electromyography (sEMG) offers non-invasive neuromuscular activity measurement.
- Reliable, real-time motion intention recognition is crucial for advanced human-machine interfaces, particularly in upper-limb prosthetics and exoskeletons.
- Current sEMG decoding methods face challenges in accuracy and real-time performance for interactive assistive applications.
Purpose of the Study:
- To present a structured framework for sEMG-based motion intention recognition in upper-limb assistance.
- To evaluate the efficacy of time-domain features and ensemble learning for decoding sEMG signals.
- To validate the proposed framework's performance in both offline and online interactive settings.
Main Methods:
- Acquisition of multi-channel sEMG signals during four distinct upper-limb motions.
- Standardized preprocessing, including denoising and segmentation, followed by extraction of eight time-domain features.
- Systematic feature subset selection and benchmarking of five supervised learning classifiers, with Random Forest optimized using K-fold cross-validation.
Main Results:
- A compact subset of time-domain features was identified as optimal for motion discrimination.
- The Random Forest classifier, optimized through K-fold cross-validation, demonstrated superior performance.
- The framework achieved an average intra-subject accuracy of 95.23% and 95.72% in online validation.
Conclusions:
- Time-domain feature fusion combined with ensemble learning provides robust and efficient motion discrimination from sEMG signals.
- The developed framework shows significant potential for real-time applications in upper-limb assistance, rehabilitation, and advanced human-machine interaction.
- This approach offers a reliable method for decoding user intentions, paving the way for more intuitive and responsive assistive technologies.
