Multi-user motion recognition using sEMG via discriminative canonical correlation analysis and adaptive...

Area of Science:

• Biomedical Engineering
• Rehabilitation Technology
• Signal Processing

Background:

• Surface electromyography (sEMG) interfaces face adoption challenges due to significant inter-individual variability in muscle signal generation.
• This variability hinders rapid adaptation for new users in rehabilitation settings, limiting the technology's potential.

Purpose of the Study:

• To develop a multi-user sEMG framework that addresses individual differences in sEMG signals.
• To enhance the adaptability of new users to sEMG interfaces for improved rehabilitation outcomes.

Main Methods:

• Proposed a multi-user sEMG framework integrating discriminative canonical correlation analysis and adaptive dimensionality reduction (ADR).
• The framework projects user-specific feature sets into a low-dimensional, uniform space to mitigate individual signal variations.
• ADR was employed to remove redundant sEMG features and boost motion recognition accuracy.

Main Results:

• Achieved an average recognition accuracy of 92.23% for 12 upper-limb movement categories in subjects with intact limbs.
• Demonstrated an average recognition rate of 90.52% in rehabilitation laboratory experiments.
• The framework effectively overcame individual differences in sEMG signals, enabling faster user adaptation.

Conclusions:

• The proposed framework provides an effective solution for rapid adaptation of new users to sEMG interfaces in rehabilitation.
• The integration of ADR significantly improves the accuracy and robustness of sEMG-based motion recognition.
• This advancement holds promise for expanding the application of sEMG in personalized rehabilitation.