摘要
为提高脑机接口系统分类的正确率,避免大量烦琐枯燥的学习训练过程,提高其适用性,提出一种改进的基于支持向量机的半监督学习算法。该算法针对在线采集的脑电数据,根据支持向量机判别函数处理未标签样本,调整分类器训练集,同时动态调整支持向量机中惩罚因子C的值,提高了分类器性能。实例分析表明该算法比传统半监督学习算法更适用于脑电信号分类。
To improve the accuracy of Brain Computer Interface system classification and to avoid long-term training process for improving its applicability,an improved semi-supervised learning method based on support vector machine （SVM） is proposed.The algorithm based on the support vector machine discriminate function of unlabeled samples of online EEG data,adjust the classifier training set,and dynamic adjusting value of the penalty factor C in SVM,improve the performance of classifier.Contrasting with the traditional semi-supervised learning algorithm,the results show that the algorithm is applicable to the EEG signal classification.
出处
《电子测量技术》
2014年第5期9-12,共4页
Electronic Measurement Technology
基金
唐山市科学技术研究与发展指导计划项目:思维脑电独立源信号提取的研究(12110210b)
基于层次分析法的群决策系统的研究(12140215a)
关键词
半监督学习
脑机接口
支持向量机
semi-supervised learning
brain computer interface
support vector machine
