Feature selection method based on Menger curvature and LDA theory for a P300 brain-computer interface.

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interface (BCI) systems enable direct communication by decoding electroencephalogram (EEG) signals.
• P300 spellers, a common BCI application, identify P300 event-related potentials for character recognition.
• Existing P300 spellers face challenges with feature redundancy and noisy signals, limiting real-world usability.

Purpose of the Study:

• To propose a novel hybrid feature selection method for P300-based BCI systems.
• To address the issue of feature redundancy and improve classification accuracy.
• To enhance the practical performance of BCI spellers.

Main Methods:

• A hybrid feature selection approach combining Menger curvature and linear discriminant analysis is proposed.
• Feature gain estimation, ranking, and criterion-based selection are applied.
• The intersection of features selected by both methods identifies an optimal subset.

Main Results:

• The proposed method was evaluated on three public datasets (BCI Competition III, BNCI Horizon, EPFL).
• Experimental results demonstrate superior or comparable performance against existing methods.
• The hybrid approach achieved the highest classification accuracy across all datasets after all epochs.

Conclusions:

• The novel hybrid feature selection method effectively reduces redundancy in P300-based BCI systems.
• This approach offers a new strategy for enhancing the performance and accuracy of BCI spellers.
• The findings suggest improved usability for P300 spellers in practical applications.