Determination of most effective rhythm, domain & stimulation technique for noninvasive Brain Computer Interface through SVM algorithm

Abstract

Brain Computer Interface (BCI), controlling of computer using brain, is a new dimension in the field of human computer interface (HCI). Stimulation, signal analysis and pattern recognition are three most important units in BCI system. But it is very confusing to select a stimulation technique, to select a domain to analysis EEG signal and to choose a definite rhythm for pattern recognition for most effective BCI system. In this paper, motor activity (eye movement), visual and audio stimulation both for time and frequency domain are analyzed to get highest recognition rate. Support Vector Machines algorithm is used to calculate accuracy of recognition. It is seen that, motor activity (Eye movement) is most dominant for BCI system with 87.5% accuracy of recognition for frequency domain analysis. Visual stimulation is better comparing to audio stimulation with 83.33% recognition rate in frequency domain. Theta and alpha rhythms, i.e., the frequency range (4–13) Hz, is main significant frequency band for pattern recognition, though it varies with different stimulations and mental status. So most effective BCI based cursor control system can be implemented by the combination of visual stimulation and motor activity through frequency domain analysis, because higher rate of recognition is possible by this combination.