Analysis of the brain's network based on electroencephalograms and classification of movement events using connectivity maps and features

The brain-computer interface (BCI) is a technology that allows individuals with physical disabilities to communicate and interact with the external environment using brain signals. One of the key applications of BCI is motor imagery (MI), where an individual imagines performing a movement, and the corresponding brain activity is captured through electroencephalogram (EEG) signals. MI based BCI systems have attracted considerable attention due to their potential to facilitate movement recovery and control of external devices solely through brain activity. In this study, a novel approach is presented for feature extraction and classification of EEG signals recorded during motor imagery tasks. Features were extracted using the correlation coefficient and covariance matrices. These features were subsequently classified using three different machine learning models, which are Feed Forward Neural Network (FFNN), Naive Bayes Classifier (NBC), and Linear Discriminant Analysis (LDA) Classifier. The proposed method was validated on Dataset 2a of BCI Competition IV, which includes EEG data from four distinct motor imagery categories: left hand, right hand, both feet, and tongue movements. The comparative analysis showed that the features derived from the covariance matrix consistently exceeded those based on the correlation coefficient matrix across all classifiers. In particular, the covariance matrix features achieved the highest classification accuracy. Thus, they demonstrated outstanding performance in terms of robustness and reliability in the classification of different motor imagery tasks. The results of this study highlight the effectiveness of covariance matrix based feature extraction methods for MI based BCI systems, providing a promising way to improve classification accuracy and system performance. This advancement holds significant potential for real world BCI applications, such as restoring motor function to paralyzed individuals, improving the quality of life for people with disabilities, and facilitating seamless human computer interaction in virtual and augmented reality environments. By leveraging the brain's neural activity, this approach can enable paralyzed individuals to regain a degree of autonomy and contribute to the broader development of assistive technologies and immersive interfaces.