Multiple Classifier System Based on User Feedback for BCI P300 Speller

초록

P300 is one of the common methods to detect the brain activity. It is relatively easy to generate the brain response. Signal of brain activity can be different as the conditions of subject such as hormone, fatigue and stress would be varied. So it is hard to detect the brain activity with single classifier. We propose a method of detecting the p300 signal adaptively by using the user feedback command. User feedback of selecting or canceling the command enables the system to adjust the weight of classifier and reflect the accuracy of classifiers to next phase. The experimental results show that the proposed method adaptively complements the weakness of single classifier.