Artificial Neural Network Based Classification Of Motor Imagery EEG Signals For Efficient Brain Computer Interface System

Abstract

Brain Computer Interface (BCI) systems are evolving rapidly these days because of its ability to restore independence to individuals with severe disabilities. BCI system measures the activity of the brain signal, analyses, interprets and converts them to possible actions. Motor Imagery (MI) based Electroencephalogram (EEG) signals are used widely for the development of robust BCI system. The challenges faced by MI based EEG signals are inter-subject variability, non-linearity, unpredictable and complexities in data acquisition from the subjects. In this paper, MI based EEG signals are used for the design of BCI system. The proposed work focuses on designing an accurate classifier based on Artificial Neural Network (ANN). The proposed model is trained, tested and validated on publicly available EEG dataset 'BCICompetition IV dataset 2a [7]. The efficiency of the system is evaluated based on the performance metrics such as accuracy, kappa, precision and recall. The experimental finding suggests that the proposed work for the classification of MI-EEG signals can be expanded for the creation of a reliable BCI system.