基于改进CSSD的运动想象脑电信号特征提取

张毅a， 梅彦玲a， 罗元b

(重庆邮电大学 a.信息无障碍工程研发中心； b.光纤通信技术重点实验室，重庆 400065)

摘要: 针对现有的共空域子空间（common special subspace decomposition, CSSD）算法在脑电信号（EEG）特征提取时，类内和类间的信号特征变化导致脑电信号特征值稳定性低、特征向量区分度差的问题，提出一种改进的CSSD特征提取方法，即基于KullbackLeibler距离的共空域子空间分解法（KLCSSD）。在传统CSSD算法的基础上利用KullbackLeibler距离，最大化类间距离而最小化类内差异，提取鲁棒性较强的EEG信号特征。实验结果表明：该算法相对于传统CSSD有较好的特征向量区分度，有效提高了脑电信号的正确识别率。

关键词: 脑电信号； 特征提取； KullbackLeibler距离； 共空域子空间分解法； 识别率

中图分类号： TP 242.6文献标志码： A

Feature Extraction of Motor Imagery in EEG Based on Improved CSSD

ZHANG Yia, MEI Yanlinga, LUO Yuanb

(a.Information Accessibility Engineering R&D Center;b.Key Laboratory of Optical Fiber Communication Technology,

Chongqing University of Posts and Telecommunications, Chongqing 400065, China)

Abstract: On the basis the current feature extraction algorithm of common special subspace decomposition (CSSD) in electroencephalograph (EEG), an improved CSSD basing on KullbackLeibler (KLCSSD) algorithm for extracting EEG feature was presented to solve the problems such as low stability of the eigenvalues and poor discriminative ability of eigenvectors caused by the signal characteristics changes of the class and categories in the EEG recognition process. The presented algorithm using KL distance based on the traditional CSSD algorithm, maximized difference of the categories and minimized difference of the class and extracted the EEG feature that having a good robustness. The experiment result shows that the improved CSSD algorithm has a better distinguish between the feature vectors than the CSSD and effectively improve the correct recognition.

Key words: EEG; feature extraction; KullbackLeibler distance; CSSD; recognition

收稿日期： 20140728

基金项目： 国家自然科学基金项目(60905066；51075420);科技部国际合作项目（2010DFA12160）.

作者简介： 张毅（1966—），男，教授，博士生导师.