基于集合经验模式分解和公共空间模式的脑电信号特征提取  

A Novel Electroencephalography Feature Extraction Based on Ensemble Empirical Mode Decomposition and Common Spatial Pattern

作　　者：张学军[1,2] 霍延 黄丽亚[1,2] 成谢锋[1,2] ZHANG Xue-jun;HUO Yan;HUANG Li-ya;CHENG Xie-feng(School of Electronic and Optical Engineering,Nanjing University of Posts and Telecommunications,Nanjing 210023,China;Nation-Local Joint Project Engineering Lab of RF Integration&Micropackage,Nanjing University of Posts and Telecommunications,Nanjing 210023,China)

机构地区：[1]南京邮电大学电子与光学工程学院,南京210023 [2]南京邮电大学射频集成与微组装技术国家地方联合工程实验室,南京210023

出　　处：《科学技术与工程》2020年第1期109-117,共9页Science Technology and Engineering

基　　金：国家自然科学基金(61271334)

摘　　要：公共空间模式(common spatial pattern,CSP)能够较好地提取运动想象任务时脑电信号的判别特性,但是其性能与大脑进行想象任务的频带相关。为了确定这样一组频带实现精确的分类,基于集合经验模式分解、FIR滤波器组以及公共空间模式算法提出了一种脑电特征提取方法。预处理去除伪迹后的信号首先经过集合经验模式算法获得多个模函数,然后选择出包含μ节律和β节律范围的分量实现信号重构,重构后的脑电信号作为带通滤波器组的输入得到若干子带信号集合,从每个子带信号中提取CSP特征,最后将提取的特征经过支持向量机(support vector machine,SVM)进行分类。运用该方法对脑-计算机接口(brain-computer interface,BCI)竞赛数据集进行分类,实验表明该方法能够自适应地提取、筛选和判别每个受试者的空间CSP特征,分类准确率达96.53%。The common spatial pattern(CSP)is known effective in extracting features from motor imagery electroencephalograms.However,its performance depends on the frequency bands that related to brain activities in association with motor imagery tasks.To determine such a set of frequency bands to acquire an accurate classification,a novel EEG feature extraction method was proposed based on ensemble empirical mode decomposition(EEMD)and common spatial pattern(CSP)combined with finite impulse response(FIR)filter bank.This method can effectively perform the autonomous extraction and selection of key individual spatial discriminative CSP feature.In the new method,the preprocessed EEG signal was decomposed into intrinsic mode functions(IMFs)by EEMD,the intrinsic mode functions containingμandβrhythms were selected to obtain reconstructed EEG signal,and the reconstructed EEG signal was further decomposed into multiple sun-band signals by FIR filter bank.Subsequently,the features were extracted from each sun-band signal in CSP algorithm.Finally,support vector machine(SVM)was used to classify CSP feature.This method was implemented on Brain-Computer Interface(BCI)competition data set,and the results reveal that the proposed method reached 96.53%in classification accuracy.

关 键 词：集合经验模式分解 公共空间模式分解 FIR滤波器组 支持向量机 

分 类 号：R318[医药卫生—生物医学工程]