Multi-pattern motor imagery recognition based on EEG features

Motor imagery is a classical method in brain-computer interaction, in which EEG features evoked by imaginary body movements are recognized and thought information is extracted. However, the traditional dual pattern with left/right hand has low efficiency of information transformation. In this paper we introduce a multi-pattern method of transformation to overcome this disadvantage. The method of two-dimensional time-frequency analysis combined with Fisher analysis was introduced to extract feature information of event related desynchronization/synchronization from multi-pattern imaginary movement evoked EEG signals of typical subjects. Then the support vector machine was used to establish double layers classifiers to identify and classify multi-pattern motor imagery. This method achieved an accuracy of 85.71% on recognition of imaginary movements of four different body parts. The result shows that the evoked EEG feature information of multi-pattern imaginary movements has obvious space specificity which can be used to recognize and extract thought information in brain-computer interaction. Therefore, it is worthy of further investigation.