关键词: 异步BCI, 二级分类器, 阈值判别, 支持向量机, 小波变换

Abstract: A two-level structure classifier is designed aiming to solve the problems, idle state detection and classification of different motor imagery tasks, in an asynchronous BCI system. The classifier integrates a threshold discrimination method with a Support Vector Machine, on the basis of EEG feature extraction by wavelet transform. When everyone imagines unilateral limb movement in the brain, μ rhythm on EEG signal will represent ERS and ERD phenomenon in its ipsilateral and contralateral motor cortex areas. However, there is no such phenomenon during the idle state of the brain. Based on this characteristic, this paper presents a wavelet energy threshold discrimination approach to detect the idle state of the brain. Subsequently, Support Vector Machine, based on the RBF kernel function and the cross-validation method, is utilized to classify the left and right hand motor imagery tasks. Experimental results show that the accuracy of the optimal task classification can reach 80.7% costing only about 3.0 s. Therefore, it is appropriate to the proposed classifier applied in an asynchronous real-time motor imagery brain-computer interface system.

Key words: asynchronous BCI, two-level structure classifier, threshold discrimination, Support Vector Machine, wavelet transform