Related Experiment Video
Updated: Aug 5, 2026
08:15
Capturing Dynamic Finger Gesturing with High-resolution Surface Electromyography and Computer Vision
Published on: March 28, 2025
Dual-Stream SPP-CNN for High-Precision sEMG Gesture Recognition in Human-Machine Interfaces
Zebin Li1,2, Gang Zhang1, Lifu Gao2,3
1Intelligent Control and Robotics Research Center, West Anhui University, Lu'an 237012, China.
Biomimetics (Basel, Switzerland)
|July 27, 2026
Summary
This study introduces a novel dual-stream CNN for surface electromyography (sEMG) gesture recognition. The DSSCNN method significantly improves accuracy and reduces variability for natural human-machine interaction.
Area of Science:
- Biomedical Engineering
- Machine Learning
- Human-Computer Interaction
Background:
- Surface electromyography (sEMG) signals offer natural control for human-machine interfaces.
- sEMG signal challenges include non-stationarity and inter-subject variability, hindering robust feature extraction and model generalization.
Purpose of the Study:
- To develop a robust deep learning framework for high-precision sEMG-based gesture recognition.
- To address the limitations of non-stationarity and inter-subject variability in sEMG signals.
Main Methods:
- A dual-stream spatial pyramid pooling convolutional neural network (DSSCNN) was proposed.
- sEMG signals were transformed into CWT spectrograms and GADF images as dual-channel input.
- A dual-stream architecture with spatial pyramid pooling (SPP) extracted spatiotemporal features.
Main Results:
- DSSCNN achieved 97.88% average accuracy (intra-subject) and 96.59% (LOSO).
- The method demonstrated low inter-subject variance.
- Real-time control of an unmanned ground vehicle (UGV) was successfully demonstrated.
Conclusions:
- The DSSCNN framework effectively enhances sEMG-based gesture recognition accuracy and robustness.
- This approach offers a promising pathway for advanced natural human-machine interaction systems.