计算机工程与应用 ›› 2008, Vol. 44 ›› Issue (36): 201-205.DOI: 10.3778/j.issn.1002-8331.2008.36.058

• 工程与应用 • 上一篇    下一篇

基于支持向量机的思维脑电信号特征分类研究

伍亚舟1,张 玲1,易 东1,吴宝明2   

  1. 1.第三军医大学 卫生统计学教研室,重庆 400038
    2.第三军医大学 大坪医院野战外科研究所五室,重庆 400042
  • 收稿日期:2008-06-30 修回日期:2008-09-01 出版日期:2008-12-21 发布日期:2008-12-21
  • 通讯作者: 伍亚舟

Classification of mental Electroencephalography signal features based on Support Vector Machines

WU Ya-zhou1,ZHANG Ling1,YI Dong1,WU Bao-ming2   

  1. 1.Department of Health Statistics,Third Military Medical University,Chongqing 400038,China
    2.Fifth Department of Daping Hospital & Research Institute of Surgery,Third Military Medical University,Chongqing 400042,China
  • Received:2008-06-30 Revised:2008-09-01 Online:2008-12-21 Published:2008-12-21
  • Contact: WU Ya-zhou

摘要: 探索一种实用的基于想象运动思维脑电的脑-机接口(BCI)方式,为实现BCI应用奠定比较坚实的理论和实验基础。对6名受试者进行三种不同时段(箭头出现2 s、1 s和0 s后提示按键)情况下想象左右手运动思维作业的信号采集实验,利用小波变换和支持向量机对实验数据进行离线处理。对三种情况下的延缓时间Δt0、Δt1和Δt2分析发现:Δt0与Δt1和Δt2之间都有显著性差别(p<0.05),而Δt1与Δt2之间没有显著差别(p>0.05);平均分类正确率分别达到68.00%、80.00%和56.67%(p<0.05);实际按键前0.5~1 s左右,想象左右手运动的思维脑电特征信号都发生了明显改变。通过合理的实验设计获取的信号有助于识别正确率的提高,为BCI系统中思维任务的特征提取与识别分类提供了新思路和方法。

Abstract: To explore a practical way of Brain-Computer Interface(BCI) based on imaging movement,and to extract features of Electroencephalography(EEG) for reflecting different thoughts by searching suitable methods of signal extraction and recognition algorithm,and to enhance recognition rate of communication for BCI system,to establish a substantial theory and experimental foundation for the application of BCI,different mental tasks based on imaging left-right hand movement from 6 subjects were studied at three different time sections (hinting keying after arrow appearing at 2 s,1 s and 0 s).Then authors used Wavelet Transform(WT) and Support Vector Machine(SVM) methods to process and analyze the off-line experimental data.Average delay time Δt2,Δt1 and Δt0 for all subjects at three different time sections were analyzed,it was discovered that there was significant difference between Δt0 and Δt2 or Δt1p<0.05),but there was no significant difference between Δt2 and Δt1p>0.05).The average results of classification rate were 68.00%,80.00% and 56.67% (p<0.05),respectively.There are obviously different features for imaging left-right hand movement about 0.5~1 s before practical action,these features have significant difference.Authors obtained higher classification rate of communication under hinting keying after arrow appearing about 1 s.These results show it is helpful to increase the correct rate by reasonable experimental design.The features extraction method proposed in this study has been proven feasible to be used as external control signals for BCI system.This study provides new ideas and methods for features extraction and classification of different mental tasks for BCI.