Optimized Backpropagation Neural Network Model For Brain Computer Interface System

Abstract

People with several disabilities face many challenges in personal communication with the external real-world environment. Brain Computer Interface (BCI) system offers promising solution for both communication and rehabilitation therapies which overcomes the difficulties faced by disabled people, especially fully non-speaking people. Noninvasive Electroencephalogram (EEG) based BCI system acts as a communication medium that translates brain-activity into commands for computer systems or other devices. EEG signals recorded from the scalp of aforementioned subjects are utilized to extract the meaningful patterns in the EEG-based BCI system. In this paper, an optimized Backpropagation Neural Network (BPN) based on Synergistic Fibroblast optimization (SFO) algorithm is proposed for the classification of EEG patterns in the design of robust BCI system. EEG signals collected from open repository database are employed to evaluate the efficiency of SFO based BPN classifier which is compared with Support Vector Machine (SVM) and conventional BPN method. Investigation of the results show that SFO based BPN method achieves highest classification accuracy compared to other conventional classifiers.