Improving classification accuracy of SSVEP based BCI using RBF SVM with signal quality evaluation

Power spectral density analysis (PSDA) and canonical correlation analysis (CCA) are two of the most widely used feature extraction methods for SSVEP based brain computer interfaces. However, these features may be contaminated by spontaneous EEG or noise. It is still a challenge to detect it with a high accuracy, especially at a short time window (TW) which is a tradeoff between accuracy and speed for brain computer interface. In this paper, we propose to combine both power spectral density analysis (PSDA) and canonical correlation analysis (CCA) for steady state visual evoked potential (SSVEP) feature extraction. One against one radial basis function support vector machine (OAO RBF SVM) is applied to classification in order to improve the short time window classification accuracy. Moreover, we present a signal quality evaluation method that cancels the decision of the RBF SVM when signal quality is low and prone to be misclassified. Making no decision could reduce the cost of making a wrong decision. Results show that our proposed method outperforms the standard CCA method in classifying SSVEP responses of five frequencies across four subjects. Approximately above 80% SSVEP classification accuracy is achieved when time window is above three seconds.