Reduce Calibration Time in Motor Imagery Using Spatially Regularized Symmetric Positives-Definite Matrices Based...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalogram (EEG) based brain-computer interfaces (BCI) require extensive subject-specific training, leading to prolonged calibration times.
• Current BCI decoding methods using EEG covariance matrices do not leverage the geometric properties of these matrices in the Symmetric Positive Definite (SPD) space, limiting performance.
• Reducing BCI calibration time is a critical challenge for practical BCI applications.

Purpose of the Study:

• To reduce calibration time for EEG-based BCIs by introducing a novel SPD-based classification approach.
• To address the limitations of SPD-based classification with small training datasets by proposing a dimensionality reduction framework.
• To improve the accuracy and efficiency of BCI systems through advanced signal processing techniques.

Main Methods:

• A new framework is proposed to transform SPD matrices into a lower-dimensional space using spatial filtering.
• The spatial filters are regularized by prior information derived from EEG channels.
• The approach was validated on small sample datasets (Dataset IVa, BCI Competition III) and further evaluated on larger datasets (Dataset IIIa, BCI Competition III; Dataset IIa, BCI Competition IV).

Main Results:

• The proposed approach achieved a mean accuracy of 86.13% and a mean kappa of 0.72 on Dataset IVa, outperforming existing methods.
• On Dataset IIIa and Dataset IIa, the method achieved mean accuracies of 92.22% and 81.21%, respectively, demonstrating robustness.
• The dimensionality reduction technique effectively overcomes the limitations of SPD-based classification in small training set scenarios.

Conclusions:

• The proposed SPD-based classification framework significantly reduces calibration time for EEG-BCIs.
• The method enhances BCI performance by effectively utilizing the geometric properties of SPD covariance matrices.
• This research offers a promising solution for more efficient and practical BCI system development.