Abstract
Brain-computer interfaces (BCIs) provide a non-muscular means of communication and control for severely paralyzed patients. Many BCIs depend on the P300 which is an exogenous event-related potential (ERP) component produced about 300 ms in response to the presentation of an infrequent, but recognized, stimulus. Although there are different EEG neuroheadsets in the market used to record the P300, not all of them are suitable for daily use due to the system cost and set-up time. The present study investigated the reliability of an affordable, low-cost, and wireless EEG device, namely Emotiv EPOC+, to record the P300 signals. Ten healthy volunteers tested a P300 speller system to spell 10 random characters. EEG data were recorded while the subjects attended to the standard P300 paradigm introduced by Farwell and Donchin in 1988. We examined the effect of using different matrix size (6x6 and 3x3), flash duration (100 and 175 ms), and colored matrix (white/gray and green/blue) on the performance of the P300 speller. The results show that the P300 amplitude is positively correlated with larger matrix size and longer flash duration. Moreover, the results show that using color (green/blue) stimuli enhanced larger P300 amplitude. Using linear discriminant analysis (LDA) classifier, the highest classification accuracy achieved was 75.9 \pm 7.22% when using 6x6 matrix, 175 ms flash duration, and green/blue stimuli condition. We conclude that such an affordable and wireless neuroheadset system can provide severely disabled people an alternative communication and control technology to be used effectively in their real-life environments.