An effective classification framework for brain-computer interface system design based on combining of fNIRS and EEG...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-computer interface (BCI) is a rapidly advancing field in neuroscience.
• BCIs facilitate human-computer communication via neural activity.
• Preprocessing of Electroencephalography (EEG) and Near-Infrared Spectroscopy (NIRS) signals is crucial for BCI systems.

Purpose of the Study:

• To develop and evaluate novel attribute weighting methods for BCI signal classification.
• To improve the accuracy of BCI systems using combined EEG and NIRS data.
• To compare the performance of different classifiers with the proposed weighting techniques.

Main Methods:

• Utilized datasets for Motor Imaginary and Mental Activity tasks incorporating EEG and NIRS signals.
• Extracted features including HbO, HbR, and EEG, processed using k-Means clustering centers based attribute weighting (KMCC-based) and k-Means clustering centers difference based attribute weighting (KMCCD-based) methods.
• Employed Linear Discriminant Analysis (LDA), Support Vector Machine (SVM), and k-Nearest Neighbors (kNN) classifiers for performance evaluation.

Main Results:

• Achieved high classification accuracy: 99.7% for Motor Imaginary (kNN, KMCCD-based) and 99.9% for Mental Activity (SVM/kNN, KMCCD-based).
• Demonstrated that the proposed weighting methods significantly enhance classification accuracy in EEG and NIRS BCI systems.
• The novel approach offers improved classifier performance with reduced processing power and ease of application.

Conclusions:

• The developed weighting methods are effective in improving BCI classification accuracy.
• The hybrid BCI approach combining EEG and NIRS shows significant potential.
• Future research could integrate these weighting methods with deep learning architectures for further performance gains.