Real-time implementation of a self-recovery EMG pattern recognition interface for artificial arms

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Signal Processing

Background:

• Electromyography (EMG) pattern classification is crucial for intuitive prosthetic control.
• Signal noise and disturbances significantly degrade EMG classification performance.
• Reliable user intent recognition is essential for advanced prosthetic functionality.

Purpose of the Study:

• To design and evaluate a real-time self-recovery EMG pattern classification interface.
• To improve the reliability of multifunctional prosthetic arm control.
• To address performance degradation caused by EMG signal disturbances.

Main Methods:

• Developed a novel self-recovery module with sensor fault detectors.
• Implemented a fast Linear Discriminant Analysis (LDA) classifier retraining strategy.
• Integrated the system on an embedded platform for real-time operation.

Main Results:

• The self-recovery EMG pattern recognition (PR) system demonstrated effective real-time performance.
• The system successfully recovered classification accuracy despite introduced signal disturbances.
• Experimental evaluation on an able-bodied subject validated the prototype's functionality.

Conclusions:

• The developed self-recovery module enhances the robustness of EMG-based prosthetic control.
• This technology shows potential for improving the clinical usability of EMG PR systems.
• The system offers reliable user intent recognition for multifunctional prosthetic applications.