文章摘要
程龙龙,邱爽,徐瑞,许敏鹏,奕伟波,明东,綦宏志,万柏坤.基于极能差与共空间模式算法的脑电信号特征增强研究[J].高技术通讯(中文),2013,23(9):980~987
基于极能差与共空间模式算法的脑电信号特征增强研究
Research on EEG feature enhancement based on extreme energy difference and common spatial pattern algorithms
  
DOI:
中文关键词: 脑电图(EEG),极能差(EED),共空间模式(CSP),想象动作,特征增强
英文关键词: electroencephalography (EEG),extreme energy difference (EED),common spatial pattern (CSP),movement imaginary,feature enhancement
基金项目:
作者单位
程龙龙 天津大学精密仪器与光电子工程学院生物医学工程系;无线移动通信国家重点实验室(电信科学技术研究院)北京 
邱爽 天津大学精密仪器与光电子工程学院生物医学工程系 
徐瑞 天津大学精密仪器与光电子工程学院生物医学工程系 
许敏鹏 天津大学精密仪器与光电子工程学院生物医学工程系 
奕伟波 天津大学精密仪器与光电子工程学院生物医学工程系 
明东 天津大学精密仪器与光电子工程学院生物医学工程系 
綦宏志 天津大学精密仪器与光电子工程学院生物医学工程系 
万柏坤 天津大学精密仪器与光电子工程学院生物医学工程系 
摘要点击次数: 3253
全文下载次数: 2794
中文摘要:
      为有效滤除颅骨、脑脊液、头皮等多层组织对脑电图(EEG)信号传导、衰减和混叠的影响,进一步增强信号特征以有利于认知脑电的特征提取、模式识别,以左右手想象动作电位实验为例分析比较了基于极能差(EED)和共空间模式(CSP)算法的空间滤波与特征增强效果。研究结果表明,两种算法皆在高维空间中通过方差判断能量总体分布进行空间滤波训练,可有效提高信噪比、明显增强信号特征和提升识别准确率,其中CSP算法运用矩阵同时对角化原理寻找投影方向,使方差的类间差别最大化,特征增强效果更优。以上结果可供有关脑认知科学研究与脑-机
英文摘要:
      In order to effectively reduce the influences of skull,cerebrospinal fluid and scalp as well as other tissues on the conduction,attenuation and mixture of electroencephalogram (EEG) signals and make further enhancement of signal features to promote feature extraction and pattern recognition for cognitive EEG,two algorithms,such as the extreme energy difference (EED) and common spatial pattern (CSP),were studied and their effects for spatial filter and feature enhancement were compared by using EEG data of the left/right hands motor imaginary potential experiments taken as examples.The studied results show that,both two algorithms make spatial filter training in the high dimension space by variance judgement of the total energy distribution,thus they can effectively improve the signal to noise rate (SNR),evidently enhance the signal features and increase the accuracy of pattern recognition.Among them,the CSP algorithm looks for the projecting direction by using the matrix diagonalization principle simultaneously to enlarge the variance difference between classes to be maximum and its performance is even more excellent than EED on feature enhancement.Above results would provide references for scientific research of brain cognition and for system design and application of brain computer interface (BCI).
查看全文   查看/发表评论  下载PDF阅读器
关闭

分享按钮