Intentions Recognition of EEG Signals with High Arousal Degree for Complex Task

This paper presents a novel electroencephalography (EEG) evoked paradigm based on neurological rehabilitation. By implementing a conceptual model "cup-and-ball" system, EEG signals in manipulating the dynamic constrained objects are generated. Based on the operational EEG signals, a method is proposed to recognize different mental intentions. Under the manipulating task with a high arousal level, common spatial patterns (CSP) is used to extract and optimize features of the EEG signals from ten participants. Quadratic discriminant analysis (QDA) is implemented on EEG signals in different dimensions to identify different EEG patterns. The cross-validation is used to make classifier adaptive to a given data set. The receiver operating characteristic (ROC) curves are presented to illustrate recognition performance. The classification effect of QDA is verified by paired t-test (P < 0.001). Based on the proposed method, the average accuracy of mental intentions is 0.9857 ± 0.0191 and the area under the ROC curve (AUC) is 0.9665 ± 0.0291. The performance of QDA is also compared with the other three classifiers such as the support vector machine (SVM), the decision tree (DT) and the k-nearest neighborhood (k-NN) rule. The results suggest that the proposed method is very competitive with other methods.