Abstract
In recent years, Brain–computer interfaces (BCIs) have been widely studied and become popular research topics on the applications of many fields. Power spectral density analysis (PSDA) and canonical correlation analysis (CCA) are two of the most widely used detection methods for SSVEP based brain computer interfaces. However, EEG signals are non-stationary, nonlinear and noisy so the recognition accuracy of a BCI usually decreases with time window getting shorter. And the length of time window is a tradeoff between recognition accuracy and operation speed for brain computer interfaces. Hence, it is an important issue to keep the brain computer interfaces with a high recognition accuracy when operated at short time window. In this study, we propose to combine both PSDA and CCA for SSVEP feature extraction in order to increase the information in the feature space. Cascade support vector machine is applied to classification so as to improve the recognition accuracy at short time window. Moreover, we present a signal quality evaluation method that cancels the decision of the classifier when signal quality is low and prone to be misclassified. A feedback alarm would be given to the user in order to increase user’s attention when data, which was prone to be misclassified, was detected by signal quality evaluation unit. Making no decision could reduce the cost of making a wrong decision so as to improve the error rate. Results show that our proposed method outperforms the standard CCA method in classifying SSVEP responses of five frequencies across four subjects. Above 80 % recognition accuracy is achieved when the time window is above three seconds.