Computer Engineering and Applications ›› 2009, Vol. 45 ›› Issue (21): 212-214.DOI: 10.3778/j.issn.1002-8331.2009.21.061

• 工程与应用 • Previous Articles     Next Articles

Study of objective visual EEG mode classification based on Hilbert-Huang transform

ZHOU Li-xia1,FAN Ying-le1,LI Yi 1,2   

  1. 1.Institute of Biomedical Engineering and Instrument,Hangzhou Dianzi University,Hangzhou 310018,China
    2.Institute of Biomedical Engineering and Instrument,Zhejiang University,Hangzhou 310027,China
  • Received:2008-04-21 Revised:2008-07-11 Online:2009-07-21 Published:2009-07-21
  • Contact: ZHOU Li-xia

基于Hilbert-Huang变换的目标注视脑电模式分类研究

周丽霞1,范影乐1,李 轶1,2   

  1. 1.杭州电子科技大学 生物医学工程与仪器研究所,杭州 310018
    2.浙江大学 生物医学工程与仪器学院,杭州 310027
  • 通讯作者: 周丽霞

Abstract: A research of automatic classification of visual objective is made.A new application of Hilbert-Huang transform method to a character matrix(6 by 6) is presented which is scanned repeadly.With the EMD method,the original signal is decomposed into several intrisic mode functions.With the Hilbert transform,energy-frequency distribution is computed to decide which character is imaged.Finally the pattern recognition method of nearest neighbors is applied to optimal classification.The experiment data from 3rd BCI competition is analyzed,and classification is made.The mean classification accuracy is up to 84.21%.Based on the Hilbert-Huang transform of EEG signals,the classification of character which is imaged can be made effectively,which will help to develop the BCI system.

Key words: Hilbert-Huang transform, Electro Encephalogram(EEG), P300, intrinsic mode function, event-relative

摘要: 研究目标注视任务,即类“思维拨号”任务下的脑电信号模式自动分类技术。基于时频分析原理,对循环扫视6×6字符矩阵所得的脑电信号进行Hilbert-Huang变换,用EMD(Empirical Mode Decomposition)方法把信号分解成数个本征模态函数(Intrinsic Mode Function,IMF);对每个IMF进行Hilbert变换,得到时频平面上的能量分布,即Hilbert谱;将各频率分量累加谱按时间变化特性作为关注与非关注某字符的特征量;最后通过K-近邻法对所视字符进行自动识别。对第3届国际脑机接口竞赛的数据进行分析测试,识别准确率达到84.21%。经Hilbert-Huang变换的时频分析方法,可以得到目标注视任务下的脑电信号有效特征,从而有助于脑机接口系统的实际应用。

关键词: Hilbert-Huang变换, 脑电信号, P300, 固有模态函数, 事件相关