“模拟阅读”脑机接口信号的核Fisher分类.pdfVIP

第32卷第2期　 　 　 　 　 　 　 　 　 　 　 　 　 　 　 　 　 中南民族大学学报(自然科学版)　 　 　 　 　 　 　 　 　 　 　 　 　 　 　 　 　 Vol.32 No.2 2013年6月　 　 　 　 　 　 　 　 　 　 　 Journal of South-Central University for Nationalities(Nat.Sci.Edition)　 　 　 　 　 　 　 　 　 　 Jun.2013 “模拟阅读”脑-机接口信号的核 Fisher分类 官金安,荆汉娜,李　 梅 (中南民族大学 生物医学工程学院,认知科学国家民委重点实验室,武汉430074) 摘　 要　 采用“模拟阅读”视觉诱发事件相关脑电位的实验模式进行了脑-机接口实验,记录受试者在靶标和非靶 标刺激下的脑电信号并进行了模式分类.首先对原始的脑电信号进行预处理、提取最优特征,然后利用核Fisher准 则对采集到的脑电信号进行分类.为了验证这种方法的有效性,同时还对信号进行了感知准则和Fisher 准则的分 类.对5名受试者测试的结果表明:核 Fisher 准则分类器的正确率分别为92.98%,88.85%,76.32%,78.33%和 95.52%,分别高出Fisher准则、感知准则约20%、10%,表明核Fisher准则对含有非线性可分的脑电信号的分类效 果比单纯的线性方法更为优越. 关键词　 脑-机接口;感知准则;Fisher准则;核Fisher准则 中图分类号　 TP312;R318　 文献标识码　 A　 文章编号　 1672-4321(2013)02-0087-05 Kernel Fisher Based EEG Classification for ″Imitating Reading″ Brain-Computer Interface Guan Jin′an,Jing Hanna,LiMei (Key Laboratory of Cognitive Science,State Ethnic Affairs Commission, College of Biomedical Engineering,South-Central University for Nationalities,Wuhan430074,China) Abstract　 This paper performed the experiment in the mode of “imitating reading” visual evoked event-related potentials for brain-computer interface,recorded subjects’ EEG respectively under the target and non- target stimulation,and then classified the signals into two categories. After raw EEG signals preprocessing and optimal features extracting,a kernel Fishercriterionwasusedforclassification. Ascomparison,perceptioncriteriaandFishercriteriawerealsoperformedinthe same data. The results from five subjects showed that the correct classification rates of kernel Fisher were 92.98%、88. 85%、76.32%、78.33%、95.52%,they are about 20% and 10% higher than Fisher criteria and perception criteria separately,and the kernel Fisher criter