基于Fisher判据时频分析的运动相关脑电特征选择及优化.pdfVIP

Vo l. 42 NQ8 第 42 卷第8 期 西安交通大学学报 2008 年8 月 JOURNAL OF XI AN JIAOTONG UNIVERSI丁Y Aug. 2008 基于 Fisher 判据时频分析的运动相关 脑电特征选择及优化 裴晓梅，郑崇勋 (西安交通大学生物阪学信息工程教育部嚣点实验室. 710049. 西安) 摘襄:提出了-种暴于 Fisher 判据的运动相关脑也特征优化选择的时频分析方法，为优化选择与 i运动意识最相关的有效脑也频率成分提供了理论依据.在此暴础上，利用具有高时频分辨碎的 Morlet 小波方法，提取与脑也特征优化过程最为匹配的特征信息，对 4 个受试者运动相关脑也模 式进行分类，平均最大分类正确率达到 87.95% ，通过最大分类正确率、最大互信息两项评价指标 比较，验证了暴于匹配追踪和 Fisher 判据时频分析的特征优化方法对改品大脑运动意识任务分类 性能的有效性.实验结果表明，该文所提出的方法可捏成用于脑机接口系统的运动相关脑电特征优 化和选择中. 关键词:特征选择a优化.;Fisher 判据;Morlet 小波洁、波;互信息 中团分类号: R318 文献标志码:A 文章蝙号: 0253-987X(2008)08-1026-05 Feature Selection and Optimization of Movement翩ReIated EEG Based on Fisher Criterion Time Frequency Analysis PEI Xiaomei. ZHENG Chongxun (Key Laboratory of Biomedical Information Engineering of Education Ministry. Xian ]iaotong University. Xian 710049 , Clúna) Abstract: A novel time…frequency analysis method based on Fisher criterion for optimization of movement伽related EEG power features is proposed to provide a theoretical guide for selecting the most relevant EEG frequency components. With Morlet wavelet filter to extract the optimized movement唰related EEG features , two classes of EEG patterns for 4 subjects are discriminated and the average maximum classification accuracy reaches to 87. 95%. 日y the two evaluation indexes , i. e. maximum classification accuracy and mutual inforrnation (M I), the effectiveness of feature optimization based on time两frequency analysis of Fisher-ratio for improving classification perform- ance is verified. The experimental results indicate the further applications of the propose met