Iterative Self-Training Based Domain Adaptation for Cross-User sEMG Gesture Recognition.

Area of Science:

• Biomedical Engineering
• Machine Learning
• Rehabilitation Technology

Background:

• Surface electromyography (sEMG) is crucial for gesture recognition in rehabilitation.
• sEMG signals are highly user-dependent, limiting model generalization.
• Existing domain adaptation methods struggle with complex physiological signals.

Purpose of the Study:

• To develop a novel domain adaptation method for cross-user sEMG gesture recognition.
• To address the limitations of current methods in handling user variability.
• To improve the applicability of sEMG recognition models to new users.

Main Methods:

• Proposed an Iterative Self-Training based Domain Adaptation (STDA) method.
• Utilized discrepancy-based domain adaptation (DDA) with Gaussian kernel distance.
• Implemented pseudo-label iterative update (PIU) for enhanced data labeling.

Main Results:

• STDA demonstrated significant performance improvements in cross-user sEMG gesture recognition.
• The method effectively supervised feature decoupling using self-generated pseudo-labels.
• Experiments on NinaPro and CapgMyo datasets validated the approach's efficacy.

Conclusions:

• The proposed STDA method offers a robust solution for user-independent sEMG gesture recognition.
• STDA enhances feature decoupling and pseudo-labeling for better cross-user adaptation.
• This work advances the practical application of sEMG in diverse user scenarios.