Preliminary Study of Virtual sEMG Signal-Assisted Classification

Area of Science:

• Biomedical Engineering
• Signal Processing
• Human-Computer Interaction

Background:

• Surface electromyography (sEMG) is crucial for analyzing user intentions across various applications.
• Current sEMG classification methods are electrode-dependent, limiting their practical use as classifiers trained on one set of electrodes do not generalize to others.
• This electrode dependency necessitates retraining for different electrode configurations, hindering real-world applicability.

Purpose of the Study:

• To propose a novel virtual sEMG signal-assisted classification method.
• To achieve electrode-independent classification for surface electromyography signals.
• To demonstrate the feasibility of using virtual sEMG signals for classification tasks.

Main Methods:

• Development of a model to generate virtual sEMG signals based on muscle activation.
• Utilizing the generated virtual sEMG signals to enable classification with fewer channels.
• Testing the electrode-independent classification capability of the proposed virtual sEMG approach.

Main Results:

• Demonstrated the feasibility of virtual sEMG signal-assisted classification.
• Showcased successful classifications using fewer sEMG channels with a pre-trained classifier without re-training.
• Validated the potential for electrode-independent classification using the proposed virtual signal generation method.

Conclusions:

• The virtual sEMG signal-assisted classification shows promise for overcoming electrode dependency.
• This approach enhances the practicability of sEMG-based applications by enabling classifier generalization.
• Further research is needed for model refinement and validation with more subjects to achieve plug-and-play functionality.