Typically, people with severe motor disabilities have limited opportunities to socialize. Brain-Computer Interfaces (BCIs) can be seen as a hope of restoring freedom to immobilized individuals. Motor imagery (MI) signals recorded via electroencephalogram (EEG) are the most convenient basis for designing BCIs as it provides a high degree of freedom. MI based BCIs help motor disabled people to interact with any real-time BCI applications by performing a sequence of MI tasks. But inter-subject variability, extracting user-specific features and increasing accuracy of the classifier is still a challenging task in MI based BCIs. In this work, we propose an approach to overcome the above-mentioned issues. The proposed approach considers channel selection, band-pass filter based common spatial pattern, feature extraction, feature selection and modeling using Gaussian Naïve Bayes (GNB) classifier. Since the optimal features are selected by feature selection techniques, it helps to overcome inter-subject variability and improves performance of GNB classifier. To the best of our knowledge, the proposed methodology has not been used for MI-based BCI applications. The proposed approach has been validated using BCI competition III dataset IVa. The result of our approach has been compared with two classifiers, namely Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM). The results prove that the proposed method provides an improved accuracy than LDA and SVM classifiers. The proposed method can be further developed to design a reliable and real-time MI-based BCI applications.
Key words: Motor imagery, brain computer interface, electroencephalography, feature extraction, feature selection, machine learning.
|
