Related Experiment Video
Updated: Feb 19, 2026
Author Spotlight: Enhancing Neurorehabilitation Through EEG, Motor Imagery, and Virtual Reality
Published on: May 10, 2024
CSP-TSM: Optimizing the performance of Riemannian tangent space mapping using common spatial pattern for MI-BCI
Shiu Kumar1, Kabir Mamun2, Alok Sharma3
1Department of Electronics, Instrumentation and Control, School of Electrical & Electronics Engineering, College of Engineering, Science and Technology, Fiji National University, Suva, Fiji; School of Engineering and Physics, Faculty of Science, Technology and Environment, The University of the South Pacific, Suva, Fiji.
A new Common Spatial Pattern (CSP) framework using Tangent Space Mapping (TSM) improves motor imagery brain-computer interface (MI-BCI) classification accuracy. This CSP-TSM method reduces classification errors compared to conventional CSP, offering a more efficient MI-BCI solution.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Classifying electroencephalography (EEG) signals for motor imagery brain-computer interfaces (MI-BCI) is challenging.
- Common Spatial Pattern (CSP) is a widely used technique for this classification task.
- Existing methods require improvement for enhanced MI-BCI performance.
Purpose of the Study:
- To develop a novel framework for classifying EEG signals in MI-BCI applications.
- To enhance the accuracy and efficiency of MI-BCI systems.
- To introduce a new method combining CSP with Tangent Space Mapping (TSM).
Main Methods:
- A single band CSP framework (CSP-TSM) was proposed, integrating Tangent Space Mapping (TSM) on covariance matrices.
- Spatial filtering was applied to bandpass-filtered MI EEG signals.
- Features from TSM and CSP variance were fused, selected using Lasso, and classified with Support Vector Machine (SVM) after Linear Discriminant Analysis (LDA).
Main Results:
- The CSP-TSM framework demonstrated improved MI EEG signal classification performance.
- Classification error rates were reduced by 3.16%, 5.10%, and 1.70% on BCI Competition datasets compared to conventional CSP.
- The proposed method showed superior results against competing techniques.
Conclusions:
- The CSP-TSM method yields promising results for MI-BCI, outperforming several existing methods.
- This framework offers reduced computational complexity compared to the standalone TSM method.
- CSP-TSM has the potential for developing advanced MI-BCI systems.
