Surface EMG pattern recognition for real-time control of a wrist exoskeleton
Zeeshan O Khokhar1, Zhen G Xiao, Carlo Menon
1MENRVA Group, School of Engineering Science, Faculty of Applied Science, Simon Fraser University, 8888 University Drive, Burnaby, BC, V5A 1S6, Canada.
Biomedical Engineering Online
|August 28, 2010
Summary
This study demonstrates that surface electromyography (sEMG) can accurately classify wrist torque levels for controlling an exoskeleton. Support Vector Machines (SVM) achieved high accuracy in real-time, enabling effective assistive device operation.
Area of Science:
- Biomedical Engineering
- Rehabilitation Technology
- Human-Computer Interaction
Background:
- Surface electromyography (sEMG) is utilized for hand gesture classification and prosthetic control.
- sEMG holds potential for controlling assistive devices for individuals with muscle weakness.
- Estimating user torque is crucial for effective force control in prosthetic and assistive devices.
Purpose of the Study:
- To develop and implement a pattern recognition system for estimating human wrist torque using sEMG signals.
- To validate the real-time control of a novel two-degree-of-freedom wrist exoskeleton prototype (WEP) using the developed system.
Main Methods:
- Collected sEMG data from four forearm muscles and wrist torque from eight volunteers using a custom rig.
- Extracted features including root mean square (rms) EMG amplitude, autoregressive (AR) model coefficients, and waveform length.
- Employed Support Vector Machines (SVM) for classifying different force intensity levels and validated the scheme in real-time with the WEP.
Main Results:
- Achieved average testing accuracies of approximately 88% for nineteen classes and 96% for thirteen classes.
- The classification and control algorithm executed in under 125 ms, demonstrating real-time capability.
- Successfully validated the real-time implementation of the sEMG-based torque classification for WEP control.
Conclusions:
- Classification of sEMG signals to differentiate wrist torque levels is feasible and effective.
- Utilizing only four sEMG channels is sufficient for accurate torque classification.
- SVM is a suitable technique for real-time sEMG classification and effective control of wrist exoskeleton devices.
