基于相位同步的运动想象分类  

Motion imaging classification based on phase synchronization

作　　者：蒋勤 张毅 刘鹏飞 JIANG Qin;ZHANG Yi;LIU Pengfei(School of Computer Science and Technology,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;Advanced Manufacturing Engineering School,Chongqing University of Posts and Telecommunications,Chongqing 400065,China;School of Optoelectronic Engineering,Chongqing University of Posts and Telecommunications,Chongqing 400065,China)

机构地区：[1]重庆邮电大学计算机科学与技术学院,重庆400065 [2]重庆邮电大学先进制造工程学院,重庆400065 [3]重庆邮电大学光电工程学院,重庆400065

出　　处：《华中科技大学学报（自然科学版）》2020年第1期48-54,共7页Journal of Huazhong University of Science and Technology(Natural Science Edition)

基　　金：国家自然科学基金青年基金资助项目(61803058);重庆市自然科学基金资助项目(cstc2018jcyjAX0385)

摘　　要：针对多通道脑电信号特征提取过程中存在的特征向量稳定性低及区分度差的问题,提出相位同步与空间位置相结合的特征提取算法.首先,利用相位锁定值(PLV)衡量不同模式下不同脑区的相关程度,通过相关性分析发现感觉运动的作用机制,并按相关程度选取特征电极;然后,采用共空间模式算法(CSP)对所选电极间的PLV进行特征提取;最后,使用支持向量机(SVM)完成运动想象分类.本方法在持续4 s和1 s的左右手运动想象分类中分别获得平均91.3%和82.7%的准确率,相较于传统CSP算法具有更高的分类准确率,需要的电极更少,能快速响应短时不连续性运动想象.Aiming at the problems of low stability and poor discrimination of feature vectors in feature extraction of multi-channel electroencephalogram(EEG) signals,a feature extraction algorithm combining phase synchronization and spatial location was proposed.Phase locking value(PLV) was adopted to measure the synchronization correlation degree between different electrodes in different motor imagery modes,the characteristic electrodes were obtained by analyzing the correlation indexes.Then,the common spatial pattern(CSP) algorithm was used to extract features of PLV in selected electrodes,and SVM for classification.The presented method achieves an average accuracy of 91.3% and 82.7% in the left-right hand classifications for lasting 4 s and 1 s dataset,respectively.The proposed method has higher classification accuracy and requires fewer electrodes than the conventional CSP algorithm,and can respond quickly to short-term discontinuity motor imagery.

关 键 词：同步化分析 运动想象 相位锁定 共空间模式 脑电信号 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]