基于多域的P300脑电信号特征提取  

Feature Extraction of P300 EEG Signal Based on Multi-domain

作　　者：闫秘 白雪梅[1] 张晨洁[1] 郭家赫 YAN Mi;BAI Xue-mei;ZHANG Chen-jie;GUO Jia-he(School of Electronics and Information Engineering,Changchun University of Science and Technology,Changchun 130022)

机构地区：[1]长春理工大学电子信息工程学院

出　　处：《长春理工大学学报（自然科学版）》2019年第5期87-92,97,共7页Journal of Changchun University of Science and Technology(Natural Science Edition)

基　　金：吉林省自然科学基金项目（20150101013JC)

摘　　要：针对目前P300脑电信号微弱及分类准确率低等问题,在已有的脑电信号特征提取方法的基础上,根据脑电信号在时域、频域以及空域都存在特征的特点,提出了一种将时域特征、频域特征和空域特征融合的脑电信号特征提取方法。首先,对数据进行预处理,去除噪声及伪迹;然后,对预处理后的数据进行分段叠加平均,求取时域能量熵作为时域特征,以小波变换后得到的近似系数为频域特征,通过ICA算法得到的混合矩阵为空域特征;最后使用SVM作为分类器,在BCI竞赛Ⅲ的字符拼写数据中对所提出的方案进行验证,结果表明,对P300脑电信号进行多域特征提取能够提高P300的分类准确率,而且进行5次叠加平均的分类结果高于单一特征提取方法进行10次特征提取的分类结果,在15次叠加平均的情况下分类准确率达到了97.3%。Aiming at the weak and low classification accuracy of P300 EEG signal,this paper proposes an EEG feature extraction method which combines time domain features,frequency domain features and spatial domain features on the basis of existing EEG feature extraction methods and according to the characteristics of EEG signal in time domain,frequency domain and spatial domain.Firstly,the data are preprocessed to remove noise and artifacts;secondly,the preprocessed data are averaged by piecewise superposition,and the energy entropy in time domain is obtained as the time domain feature,the approximate coefficients obtained by wavelet transform are the frequency domain feature,and the mixed matrix obtained by ICA algorithm is the spatial domain feature.Finally,SVM is used as classifier to validate the proposed scheme in the character spelling data of BCI Competition Ⅲ.The results show that multi-domain feature extraction of P300 EEG signal can improve the classification accuracy of P300,and the average classification result of 5 times superposition is higher than that of 10 times feature extraction by single feature extraction method,and the classification accuracy reaches 97.3%under 15 times superimposed average.

关 键 词：P300脑电信号 特征提取 时域能量熵 小波变换 ICA 

分 类 号：R318.04[医药卫生—生物医学工程]