Improving classification accuracy of SVM ensemble using random training set for BCI P300-speller

P300-speller is a communication system based on brain-computer interface (BCI) which allows users to input characters by focused attention. Support vector machine (SVM) ensemble has been successfully applied to classify in the P300-speller. However, large scale of training data was needed for the sub-classifiers to ensure the performance of classification. Subjects gradually got fatigue during the period of training data collection, which resulted in an obvious difference of EEG data from early to later and affected the performance of classification. In present study, we adopted a random selection strategy to construct the training set for SVM sub-classifiers in a BCI P300-speller with familiar face paradigm. Our results showed that the random selection strategy improved the classification accuracy of SVM ensemble in BCI P300-speller.