Related Experiment Video
Updated: Aug 11, 2026
10:14
Motor Imagery Performance Through Embodied Digital Twins in a Virtual Reality-Enabled Brain-Computer Interface Environment
Published on: May 10, 2024
Contrastive Learning Network based on Multi-Scale Transformer (CLMT-net): EEG decoding for fine-grained motor imagery
Lei Zhu1,2, Qifeng Yue1, Aiai Huang1
1School of Automation, Hangzhou Dianzi University, Hangzhou, China.
Computer Methods in Biomechanics and Biomedical Engineering
|August 10, 2026
Summary
This study introduces a novel Contrastive Learning Network based on a Multi-Scale Transformer (CLMT-Net) for decoding motor imagery (MI) brain signals. CLMT-Net enhances brain-computer interface (BCI) accuracy for fine-grained movements, particularly within the same limb.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Machine Learning
Background:
- Brain-computer interfaces (BCIs) utilize electroencephalography (EEG) signals for control.
- Decoding fine-grained motor imagery (MI) within the same limb is difficult due to similar neural patterns.
Purpose of the Study:
- To develop an advanced method for accurate fine-grained motor imagery decoding.
- To improve the performance of BCIs in distinguishing subtle, same-limb movements.
Main Methods:
- Proposed a Contrastive Learning Network based on a Multi-Scale Transformer (CLMT-Net).
- Integrated multi-scale temporal convolution, FFT-based frequency fusion, spatial convolution, and dual-path Transformer.
- Employed supervised contrastive learning to enhance feature discrimination.
Main Results:
- CLMT-Net achieved a decoding accuracy of 76.13% ± 6.77% on the MI-2 dataset.
- Demonstrated competitive performance for challenging same-limb MI decoding tasks.
- The 95% confidence interval was [73.33, 78.92].
Conclusions:
- CLMT-Net effectively learns complementary EEG representations for improved MI decoding.
- The proposed method shows significant potential for enhancing BCI applications requiring precise motor control.
