Related Experiment Video
Updated: Mar 28, 2026
08:15
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
1.4K
Adversarially regularized transformer with channel-wise noise for robust hand gesture recognition using surface
Doheun Cha1, Dong-Gyu Lee2, Sangtae Ahn1
1School of Electronic and Electrical Engineering, Kyungpook National University, Daegu, 41566 South Korea.
Biomedical Engineering Letters
|March 27, 2026
Summary
This study introduces a new network for surface electromyography (sEMG) hand gesture recognition, enhancing accuracy and robustness against noise and attacks. The method shows state-of-the-art performance in prosthetic control and human-computer interaction applications.
Area of Science:
- Biomedical Engineering
- Machine Learning
- Signal Processing
Background:
- Surface electromyography (sEMG) enables hand gesture recognition via skin surface electrical signals.
- sEMG signal non-stationarity and noise hinder reliable classification.
- Existing methods struggle with robustness and generalization.
Purpose of the Study:
- To develop a novel network for robust sEMG-based hand gesture recognition.
- To improve classification accuracy and generalization despite signal challenges.
- To enhance model resilience against adversarial attacks.
Main Methods:
- Proposed a novel network integrating adversarial training with noise perturbation.
- Utilized benchmark datasets NinaPro DB1, DB2, and DB4 for evaluation.
- Tested performance in aggregated and subject-wise evaluations.
Main Results:
- Achieved state-of-the-art results on NinaPro DB1, DB2, and DB4 datasets.
- Demonstrated significant performance improvements compared to EMGHandNet.
- Showcased enhanced robustness against Fast Gradient Sign Method (FGSM) and Projected Gradient Descent (PGD) attacks.
Conclusions:
- The proposed network significantly enhances sEMG hand gesture recognition accuracy and robustness.
- Adversarial training proves effective in improving generalization and resilience.
- The approach addresses critical limitations for real-world applications like prosthetic control.
