基于融合规则的注意相关脑电分类算法研究  

A New Classifier for the Recognition of EEG Patterns Associated to Attention Tasks

作　　者：刘明宇[1] 王珏[1] 杨琴[1] 郑崇勋[1] 

机构地区：[1]西安交通大学生命科学与技术研究所,教育部生物信息与工程重点实验室

出　　处：《北京生物医学工程》2006年第4期388-391,共4页Beijing Biomedical Engineering

基　　金：国家自然科学基金(60271025);陕西省科技厅(2003K10-G24)资助

摘　　要：提出一种新的基于支持向量机(supervectormachine,SVM)学习机制和数据融合理论的脑电分类算法,并设计了注意分级实验进行验证。首先,对脑电信号进行3级小波分解,由主分量分析(principlecomponentanalysis,PCA)方法提取其中的主特征分量;然后由支持向量机对特征分量进行分类;最后依据数据融合理论,对多导分类结果进行综合判断。结果表明,该方法具有良好的鲁棒性,对多导注意相关EEG的分类准确率可达89%左右,并高于单导最优准确率,对注意力缺陷反馈治疗、注意力机制研究等有较高的实用价值。A novel algorithm based on Support Vector Machine （SVM） and fusion rule to classify the attention-related Electroencephalogram is proposed. Wavelet Transform and Principle Component Analysis （PCA） are used to pick up main features vectors from raw EEG. Then the feature vectors are classified with SVM. After that classification result is estimated synthetically. Then the feathure vectors are classified with SVM. After classification result is estimated synthetically. The results indicate the classification accuracy of the algorithm with fusion rule reachs about 89 percent and is better than that of the best channel. At the same time, selecting suitable kernel functions can obtain better result. The experiments show the algorithm is a powerful tool for cure and research EEG patterns associated to attention tasks.

关 键 词：支持向量机 数据融合理论 脑电 小波 

分 类 号：R318.04[医药卫生—生物医学工程]