作者投稿和查稿 主编审稿 专家审稿 编委审稿 远程编辑

计算机工程 ›› 2013, Vol. 39 ›› Issue (6): 255-260. doi: 10.3969/j.issn.1000-3428.2013.06.057

• 人工智能及识别技术 • 上一篇    下一篇

基于K近邻互信息估计的EEG伪迹消除方法

何海洋,罗志增   

  1. (杭州电子科技大学机器人研究所,杭州 310018)
  • 收稿日期:2012-05-29 出版日期:2013-06-15 发布日期:2013-06-14
  • 作者简介:何海洋(1984-),男,硕士研究生,主研方向:人工智能,模式识别;罗志增,教授
  • 基金资助:
    国家自然科学基金资助项目(61172134)

EEG Artifacts Removal Method Based on K-nearest Neighbors Mutual Information Estimation

HE Hai-yang, LUO Zhi-zeng   

  1. (Robot Research Institute, Hangzhou Dianzi University, Hangzhou 310018, China)
  • Received:2012-05-29 Online:2013-06-15 Published:2013-06-14

摘要: 针对脑电信号中的眼电和心电串扰伪迹,提出一种基于最小相依成分分析的互信息(MILCA)算法的伪迹消除方法。在提升小波硬阈值法对多路原始脑电信号去噪基础上,运用MILCA算法对各通道信号进行盲源分离,同时采用信号间互相关系数和互信息量作为指标,分析伪迹分离程度。与Extend Infomax、FastICA 2种常见盲源分离算法的对比结果表明,运用MILCA算法对脑电信号中的眼电及心电伪迹的分离结果最理想。

关键词: 脑电信号, 盲源分离, K近邻, 互信息, 最小相依成分分析

Abstract: In order to eliminate artifacts emanating from ocular and cardiac activities in the recorded Electroencephalogram(EEG) signals, a blind source separation algorithm of Mutual Information(MI) based on Least dependent Component Analysis(MILCA) is presented. The lifting wavelet hard-threshold method is used to denoise the multi-channel signals from original EEG, then it applies MILCA algorithm to process the EEG signals. Cross-correlation coefficient and overall MI are used as performance index to evaluate separation effects. Compared with Extend Infomax and FastICA two common blind source separation algorithms, results indicate that the MILCA algorithm is more available to remove artifacts in contaminated EEG signals.

Key words: Electroencephalogram(EEG), Blind Source Separation(BSS), K-nearest neighbors, Mutual Information(MI), least dependent component analysis

中图分类号: