Classification of ECoG motor imagery tasks based on CSP and SVM

This paper mainly studies about the data processing of brain computer interface(BCI) and presents a kind of method for classifying the ECoG motor imagery tasks. Both the training and testing ECoG datasets were filtered with the frequency band of 8-30Hz according to the event-related desynchronization and synchronization(ERD/ERS) phenomenon. The features were extracted by using Common Spatial Pattern(CSP) and then the best starting and ending points were decided through a ten-fold cross validation(CV) of the training dataset. All the features were fed into a linear support vector machine(SVM), and the final classification accuracy of testing dataset is 90%.