基于SVM_RFE的多任务导联选择算法建模  

Multi-tasks Channel Selection Algorithmic Modeling Based on SVM-RFE

作　　者：冯建奎 金晶[1] 王蓓[1] 牛玉刚[1] 王行愚[1] Feng Jiankui;Jin Jing;Wang Bei;Niu Yugang;Wang Xingyu(College of Information Science and Engineering,East China University,Shanghai 200237,China)

机构地区：[1]华东理工大学信息科学与工程学院,上海200237

出　　处：《系统仿真学报》2018年第12期4506-4512,共7页Journal of System Simulation

基　　金：国家重点研发计划(2017YFB13003002);国家自然科学基金(61573142;61773164;91420302);111引智计划(B17017)

摘　　要：在BCI (Brain computer interface)的研究中,导联选择能够用于确定与目标任务关联较大的脑功能区域。以往的导联选择方法都是基于一个数据集进行统一的导联选择,不同任务下的对应脑区是不一样的,无法抑制同一数据集中不同任务特征的干扰。基于运动想象脑电数据展开相关研究,利用SVM_RFE (Support vector machine recursive feature elimination)导联选择方法以召回率作为依据为两类运动任务分别选择最适合的导联。研究结果表明,基于SVM_RFE多任务导联选择方法在平均分类准确率上优于传统的SVM_RFE导联选择方法。Channel selection is used to locate the mental task related brain area in the brain computer interface systems.In the previous studies,the channels were selected based on the data recoded from the multi-mental tasks.However,the different mental tasks are corresponding to the different brain area.If one same brain area is selected for different mental tasks,the features of one mental task would be noises of another mental task.In this paper,a new method was presented to solve this problem based on the data of motor imagery tasks.The SVM-RFE method was used to select the channels for each motor imagery task based on recall rate.The result showed that the proposed channel selection method was superior to the traditional method in classification accuracy.

关 键 词：脑机接口 运动想象 SVM_RFE 多任务 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]