Related Experiment Video
Updated: Aug 8, 2026
10:14
Motor Imagery Performance Through Embodied Digital Twins in a Virtual Reality-Enabled Brain-Computer Interface Environment
Published on: May 10, 2024
ReCIL: Rehearsal-Based Class Incremental Learning for Cross-Subject Motor Imagery Classification
IEEE Transactions on Bio-Medical Engineering
|August 6, 2026
Summary
This study introduces a new method for brain-computer interfaces that allows motor imagery (MI) models to learn new classes sequentially without retraining. This approach improves efficiency and data privacy in brain-computer interface (BCI) applications.
Area of Science:
- Neuroscience
- Machine Learning
- Biomedical Engineering
Background:
- Electroencephalography (EEG) based motor imagery (MI) is a key brain-computer interface (BCI) technology for neuro-rehabilitation and device control.
- Current MI classification methods require retraining for new classes, which is inefficient and impractical due to data loss or privacy concerns.
Purpose of the Study:
- To develop a novel class incremental learning approach for cross-subject MI classification.
- To enable sequential learning of new MI classes without the need for complete data recollection and model retraining.
Main Methods:
- Proposes Rehearsal-based Class Incremental Learning (ReCIL) for cross-subject MI classification.
- Employs Euclidean Alignment to mitigate inter-subject EEG data distribution shifts.
- Utilizes global-local replay for knowledge preservation and dimensionality reduction for enhanced similarity computation.
Main Results:
- ReCIL demonstrates a balanced performance between plasticity (learning new classes) and stability (retaining old knowledge).
- Experiments on public MI datasets validate the effectiveness of the proposed method.
Conclusions:
- Presents the first cross-subject class incremental learning framework for MI classification.
- Offers a practical solution for expanding MI classification capabilities incrementally without starting from scratch.
