基于运动想象的脑电信号的分类研究.pdfVIP

文章编号：1001-9944(2008)04-0001·04 基于运动想象的脑电信号的分类研究 廖彩萍，李远清，赵 慧 (华南理工大学 自动化科学与工程学院，广州510640) 摘要：对基于左右手想象的脑机接口数据进行了离线分析。在预处理过程中，采用4阶椭圆带通 滤波器进行8~12Hz带通滤波．提取Mu节律对应的EEG信号。选取时间为4-5，5s内的C3、C4通道 的幅值和之比作为时城分类特征；在频域范围内．对滤波后数据进行10阶的AI(模型功率谱估 计作为频域特征：将时域特征和频域特征结合成时频特征向量作为分类特征：在分类器的选择 上．使用线性感知器作为分类准则对特征向量进行分类训练。结果表明．经过滤波后的分类结 果比未经过滤波的效果要好。选用C3、C4通道的时频特征向量作为分类特征。表达意义简单、 明了。且能将某些时频的优点结合．分类准确率较高，且分类速度快，能满足实时性要求。 关键词：脑电信号；脑机接口；时频特征；带通滤波；线性感知器 中图分类号：TP274 文献标志码：A Study of Classification of EEG Based on Motor Imageries LIAO Cai—ping，LI Yuan—qing，ZHAO Hui (Sehod of Automatic Science and Engineering，South China University of Technology，Guangzhou 510640，China) Abstract：In this paper，outline analysis to BCI data which is based on motor imagination of left and right hand is finished。In the preprocessing，we use the 4-order band-pass elliptical filter to do 8~12Hz filtering in EEG．We extract the amplitude ra- tio of C3 and C4 in 4~5．5 second as time feature，In frequency extraction．we use the 10一order AR model power spectral den- sity as to be the frequency feature．And time—frequency feature combines the time feature and the frequency feature as the fi- nal feature vector．In this paper，we choose the pereeptron algorithm as the classifier to classify the time-frequency feature vector，From the classified result it can be seen that the effection after filtering is better than before．Selecting the time—fre— quency feature vector of ehannd of C3，C4 as the classified feature makes it clearly and simply，and Can combines the advan- tages of time feature and frequency feature，and the classification is great and fast，and can meet the need of timing． Key word：electroencephalogram(EEG)；brain-computer interface(BCI)；time～~equency feature；band—pass