Increasing the robustness against force variation in EMG motion classification by common spatial patterns.

Area of Science:

• Biomedical Engineering
• Rehabilitation Engineering
• Signal Processing

Background:

• Myoelectric prostheses rely on electromyography (EMG) pattern recognition for control.
• Variations in force levels during motion alter EMG patterns, degrading classifier performance.
• Existing classifiers struggle with the variability of EMG signals due to force changes.

Purpose of the Study:

• To investigate the effectiveness of Common Spatial Pattern (CSP) features for enhancing EMG pattern recognition robustness against force variation.
• To compare the performance of CSP features against traditional time-domain (TD) features in myoelectric control.
• To improve the reliability and accuracy of EMG-based myoelectric prostheses.

Main Methods:

• Acquired EMG signals from three able-bodied subjects performing motions at low, medium, and high force levels.
• Extracted CSP features from EMG signals for motion classification.
• Compared classification accuracies of CSP features versus TD features.

Main Results:

• CSP features demonstrated superior robustness against force variation compared to TD features.
• Average classification accuracy increased by 5.3% using CSP features.
• At low force levels, CSP features achieved 84.2% accuracy, an 18.5% improvement over TD features.

Conclusions:

• CSP features offer a promising approach to increase the robustness of EMG-based myoelectric control.
• The findings suggest CSP can mitigate performance degradation caused by force variations in myoelectric prostheses.
• Further research into CSP for advanced prosthetic control is warranted.