Classification of ECoG signals for motor imagery tasks

The electrocorticogram(ECoG) is proved to have high signal-to-noise ratio(SNR), which makes it better fitting for BCIs. And this paper represents a kind of classification method of ECoG signals for motor imagery tasks(left finger and tongue). Band power(BP) with the frequency band of [8 30] was extracted as the feature, and the linear discriminant analysis(LDA), k-nearest neighbor(kNN) rules and linear support vector machine(SVM) were used as the classifiers. From the results of these three classifiers, kNN with k=7 performed better than all the other classifiers, and the classification accuracy was 87%. But the combination of these three classifiers could improve the final results a little better, which could be up to 89%.