摘要
在脑电图(EEG)信号识别中,EEG信号的采样环境、病人状态的多样性导致分类器训练所用的源域与分类器测试所用的目标域不匹配,分类器在目标域上表现不佳。为此,引入邻域适应策略,提出一种基于子空间相似度的改进主成分分析特征提取方法(SSM-PCA),在选择主成分时,考虑源域和目标域数据的几何和统计特性,并结合迁移学习分类器大间隔投射迁移支持向量机(LMPROJ),给出以SSM-PCA为基础的LMPROJ分类识别方法。实验结果表明,与结合PCA特征抽取技术和K近邻分类器实现的识别方法相比,该方法在识别正确率方面得到较大提升。
Many practical applications for epilepsy detection,the diversity of the health status of epilepsy patient and the timing of Electroencephalogram（ EEG） signal measurements lead to the mismatching between the source domain used for classifier trained and target domain used for testing. The classifiers usually do not perform well on the target domain. In order to overcome this shortcoming,an improved Principal Component Analysis（ PCA） feature extraction method called Subspace Similarity Measure Based Principal Component Analysis（SSM-PCA） is proposed,and a new classification method,named as SSM-PCA-LMPROJ is proposed by integrating SSM-PCA and the classical classifier Large Margin Projected Transductive Support Vector Machine （ LMPROJ ） . Experimental results show that the proposed method has obvious advantages compared with the traditional method,such as the method combining PCA and K Nearest Neighbor （ KNN） classifier.
出处
《计算机工程》
CAS
CSCD
北大核心
2015年第6期158-164,共7页
Computer Engineering
基金
国家自然科学基金资助项目(61170122)
教育部新世纪优秀人才支持计划基金资助项目(NCET-12-0882)
关键词
特征迁移
迁移学习
脑电图信号
特征提取
分布多样性
主成分分析
feature migration
migration learning
Electroencephalogram （EEG） signal
feature extraction
diversityof distribution
Principal Component Analysis（PCA）
