一种基于参数优化C-SVM的脑电信号分类方法及应用

发布时间：2018-03-08 16:44

本文选题：支持向量机(SVM)　切入点：脑电信号(EEG)　出处：《重庆邮电大学学报(自然科学版)》2014年01期 　论文类型：期刊论文

【摘要】：脑电信号(electroencephalograph,EEG)由于自身信号微弱且容易受到周边环境和大脑内部其他活动的影响,对其进行特征分类并提高分类准确率这一问题一直是脑机接口领域的难点。传统的基于支持向量机(support vector machines,SVM)的脑电信号特征分类方法在选取惩罚参数与核函数参数时大都只是采用经验数据,而忽略了参数优化对提升SVM分类效果重要性,而现有的参数优化方法计算复杂严重影响了分类效率。针对以上问题,提出了一种通过交叉检验和LOO误差上界对C-SVM中的惩罚参数C和核函数参数进行优化的方法,并在理论分析的基础上结合实验证明了参数优化后的分类方法能够有效提高脑电信号分类的准确率且对分类效率影响不大。
[Abstract]:Electroencephalography (EGG) is an electroencephalography (EEG) signal that is weak and vulnerable to the effects of its surroundings and other activities within the brain. The problem of feature classification and improvement of classification accuracy is always a difficulty in the field of BCI. The traditional feature classification method of EEG based on support vector machine support vector machines is used to select penalty parameters and kernel function parameters. Most of the time is based on empirical data. The importance of parameter optimization in improving the classification effect of SVM is ignored, and the complexity of existing parameter optimization methods seriously affects the classification efficiency. A method of optimizing penalty parameter C and kernel function parameter in C-SVM by cross test and LOO error upper bound is proposed. On the basis of theoretical analysis and experiments, it is proved that the optimized classification method can effectively improve the accuracy of EEG classification and has little effect on the classification efficiency.
【作者单位】：重庆邮电大学智能系统及智能机器人研究所;
【基金】：科技部国际合作项目(2010DFA12160)~~
【分类号】：TP18;R318

【参考文献】