Related Experiment Video
Updated: Feb 26, 2026
09:42
Author Spotlight: Using Motor Imagery Brain-Computer Interface to Improve Motor and Cognitive Function in Stroke Patients
Published on: September 1, 2023
2.2K
Current Source Density Estimation Enhances the Performance of Motor-Imagery-Related Brain-Computer Interface
Summary
The current source density (CSD) method enhances motor-imagery brain-computer interface performance compared to common referencing. Using more EEG channels during preprocessing improves classification accuracy for all Laplacian methods.
Area of Science:
- Neuroscience
- Biomedical Engineering
- Signal Processing
Background:
- Brain-computer interfaces (BCIs) enable communication and control through brain signals.
- Motor-imagery (MI) tasks are crucial for BCI applications, requiring accurate electroencephalography (EEG) signal processing.
- EEG referencing and surface Laplacian methods significantly influence BCI performance.
Purpose of the Study:
- To evaluate the impact of different EEG referencing schemes and spherical surface Laplacian (SSL) methods on MI-BCI classification accuracy.
- To compare common referencing (CR) and common average referencing (CAR) alongside current source density (CSD), finite difference, and SSL methods.
- To investigate the influence of channel count during preprocessing on classification performance.
Main Methods:
- EEG signals from MI tasks were preprocessed using two referencing schemes (CR, CAR) and three Laplacian methods (CSD, finite difference, SSL).
- A filter bank common spatial pattern (FB-CSP) algorithm was employed for feature extraction.
- Support vector machine (SVM) was used for both binary and four-class classification of MI tasks.
Main Results:
- The CSD method demonstrated superior performance over CR, yielding improvements of 3.02% in binary and 5.59% in four-class MI classification.
- Utilizing a larger number of channels during preprocessing consistently improved classification accuracy across all Laplacian methods.
- A significant reduction in classification efficiency was observed when fewer channels were considered during preprocessing for all surface Laplacian methods.
Conclusions:
- The CSD method offers a significant advantage for MI-BCI classification compared to common referencing.
- Optimizing the number of channels in the preprocessing stage is critical for maximizing classification accuracy in EEG-based BCIs.
- The choice of EEG referencing and surface Laplacian method, along with channel selection, critically impacts the performance of motor-imagery BCIs.
