Single-trial motor imagery electroencephalogram intention recognition by optimal discriminant hyperplane and...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Spatial filtering is crucial for enhancing electroencephalogram (EEG) signals in brain-computer interface (BCI) systems.
• Existing methods for EEG feature extraction can be improved for greater accuracy and interpretability.

Purpose of the Study:

• To propose and evaluate a novel spatial domain filtering method for EEG feature extraction called optimal discriminant hyperplane-common spatial subspace decomposition (ODH-CSSD).
• To extract and optimize multi-dimensional EEG features for improved BCI performance.
• To assess the effectiveness of the interpretable discriminative rectangular mixture model (DRMM) for classifying these optimized features.

Main Methods:

• Extracted multi-dimensional EEG features using common space subspace decomposition (CSSD).
• Established an optimal feature criterion to identify a multi-dimensional optimal projection space.
• Defined a cost function based on discriminant criterion extreme values to solve for orthogonal discriminant vectors.
• Projected EEG features into the N-dimensional optimal feature space.
• Utilized the discriminative rectangular mixture model (DRMM) for feature identification and classification.

Main Results:

• The ODH-CSSD method achieved high accuracy (over 0.91, reaching 0.975) in identifying two-dimensional optimal EEG features from motor imagery datasets.
• DRMM demonstrated the most stable recognition accuracy for two-dimensional features, with an average clustering accuracy of 0.942, significantly outperforming FCM and K-means.
• DRMM generally showed higher clustering accuracy than FCM and K-means for three-dimensional optimal features.
• The optimization of EEG features via optimal projection surpassed Fisher's ratio.

Conclusions:

• The proposed ODH-CSSD method effectively optimizes multi-dimensional EEG features for BCI applications.
• The DRMM provides an interpretable and accurate classification approach for the optimized EEG features.
• This study offers a valuable alternative for EEG feature extraction and application in BCI systems.