The Fuzzy Pattern Recognition for Motor Imagery EEG

Abstract

Motor Imagery Electroencephalography (MI-EEG) is a nonlinear signal with multiscale and fuzzy characteristics and its recognition has good applications for the rehabilitation of stroke or hemiplegia patients. In the previous study, few of experts develop fuzzy recognition methods from the point of MI-EEG signal characteristics. In this paper, the improved Multiscale Fuzzy Entropy (IMFE) is integrated with Fuzzy Support Vector Machine (FSVM) based on Fuzzy C-Means (FCM) to yield a novel fuzzy recognition method, denoted as IF-FSVM. IMFE is applied to extract the features of MI-EEG signals. And the fuzzy memberships of the features are calculated by using FCM. Furthermore, FSVM is employed as a classifier to recognize the pattern of movement imagination for MI-EEG. Experiments are performed on a public BCI competition dataset, and the results of 10-fold Cross Validation show that IF-FSVM yields relatively higher classification accuracy compared with some common used recognition methods.