Classification for the P300-based Brain Computer Interface (BCI)

Abstract

An EEG-based brain computer interface (BCI) employs a system for immediate communication between human and computer using brain activity. An important application is a spelling device to help severely disabled persons with communication. The P300 speller aims at helping individuals unable to activate muscles to spell words by means of their bio-signals. This paper work focused on classifying P300 evoked potentials that are flashing from human brain activity. In other words, this paper represents an EEG based BCI designed for automatic classification of P300 response or component. In this paper, the performances of three classifiers are evaluated and compared in terms of different metrics. These classifiers are Support Vector Machine (SVM), Bagged Tree classifier and K-Nearest Neighbor (KNN). The distinguished accuracy using SVM, Bagged Tree classifier and KNN is S3.3%, S9.2% and 96.4% respectively. The performance evaluation results indicate that the K-Nearest Neighbor (KNN) classifier achieves the best accuracy.