Journal of South China University of Technology (Natural Science Edition) ›› 2008, Vol. 36 ›› Issue (5): 101-105.

• Electronics, Communication & Automation Technology • Previous Articles     Next Articles

Frequency-Domain Pattern Filtering of Electrocorticography Based on Motor Imagery

Lü Jun  Xie Sheng-li    

  1. School of Electronic and Information Engineering,South China University of Technology,Guangzhou 510640,Guangdong,China
  • Received:2007-04-23 Revised:2007-10-14 Online:2008-05-25 Published:2008-05-25
  • Contact: 吕俊(1979-),男,博士生,主要从事机器学习、脑机接口方面的研究. E-mail:rylj@163.com
  • About author:吕俊(1979-),男,博士生,主要从事机器学习、脑机接口方面的研究.
  • Supported by:

    国家自然科学基金重点项目(U0635001);国家自然科学基金资助项目(60505005)

Abstract:

In order to improve the classification accuracy of the brain-computer interface(BCI) of electrocortico-graphy(ECoG),a motor imagery-based pattern-filtering method in frequency domain is proposed.In this method,the joint diagonalization is employed to seek the most discriminative projections as the frequency-domain filters,the means and standard deviations of filtered electrocorticograms are extracted as the features,and the kernel Fisher discriminant is applied to the classification.Experimental results of BCI2005 data set Ⅰ show that the proposed method can achieve a classification accuracy of 92% even with a single electrode.

Key words: electrocorticography, brain-computer interface, feature extraction, frequency-domain pattern filtering