An efficient P300-speller for Arabic letters

Brain-computer interfaces (BCI) aim at enabling the brain to directly control an artificial device. One of the most popular BCI paradigms is the so called ‘P300 speller’ whose aim is to offer patients unable to speak and neither move to communicate. In this work, we present a ‘P300 speller’ developed for Arabic letters for the first time. We propose a new automatic algorithm based on the independent component analysis (ICA) to remove eye blinks artifacts, and an adaptive channel selection procedure is also proposed. Results show that the classification performance of the ‘P300-speller’ increase significantly using these two proposed techniques. Four classification algorithms were compared for offline classification from eleven participants. The results indicate that while all methods attained acceptable performance levels, the classical Support Vector Machine (SVM) classifier provide the best performance for classification of P300-speller.