Computer Engineering and Applications ›› 2017, Vol. 53 ›› Issue (13): 9-15.DOI: 10.3778/j.issn.1002-8331.1703-0199

Previous Articles     Next Articles

EEG signals feature extraction combined with empirical mode decomposition and common spatial pattern

ZHANG Xuejun1,2, HUANG Wanlu1, HUANG Liya1,2, CHENG Xiefeng1,2   

  1. 1.School of Electronic Science and Engineering, Nanjing University of Posts and Telecommunications, Nanjing 210023, China
    2.Jiangsu Province Engineering Lab of RF Integration & Micropackage, Nanjing 210023, China
  • Online:2017-07-01 Published:2017-07-12


张学军 1,2,黄婉露1,黄丽亚1,2,成谢锋1,2   

  1. 1.南京邮电大学 电子科学与工程学院,南京 210023
    2.江苏省射频集成与微组装工程实验室,南京 210023

Abstract: Normal Common Spatial Pattern(CSP) method is restricted to the abundant input channels and lacking frequency information. This paper puts forward an improved CSP method combined with the Empirical Mode Decomposition(EMD-CSP) to achieving feature vector by a different component choice. Firstly, the EMD method is proposed to decompose the EEG signal into a set of stationary time series called Intrinsic Mode Functions(IMF). Secondly, these IMFs are analyzed with the band-power to detect the valuable IMFs with characteristics of sensorimotor rhythms(5~28 Hz), and then the improved CSP filter is attached to the feature extraction of screening IMFs. Finally, once the feature vector is built, the classification of MI is performed using Support Vector Machine(SVM). The results obtained show that the EMD-CSP allow the most reliable features and that the accurate classification rate obtained is 92% which confirms the feasibility and availability of this method.

Key words: Electroencephalogram(EEG), Empirical Mode Decomposition(EMD), Common Spatial Pattern(CSP)

摘要: 常规的公共空间模式分解方法需要大量的输入通道、缺乏频域信息,发展受到限制。为了克服以上缺点,将经验模式分解(Empirical Mode Decomposition,EMD)和公共空间模式算法结合,改变CSP滤波器成分选择方式,提出EMD-CSP算法来获取特征向量。该算法对预处理后的信号进行经验模式(EMD)分解,得到固有模态函数(Intrinsic Mode Functions,IMFs),观察并计算每个IMF分量的能量谱,筛选有效的IMF频段(5~28 Hz),使用改进的CSP滤波器进行滤波获取特征,最后使用支持向量机(Support Vector Machine,SVM)进行分类。分类结果得到9位受试的想象运动平均分类正确率为92%,证实了该算法的可行性与有效性。

关键词: 脑电信号, 经验模式分解, 公共空间模式分解