基于因子分析模型的噪声稳健脑电信号分类方法
投稿时间:2019-07-16  修订日期:2019-08-19  点此下载全文
引用本文:
摘要点击次数: 71
全文下载次数: 0
作者单位邮编
彭锦强* 惠州城市职业学院 广东 惠州 516025 516025
中文摘要:脑电信号的非线性、非平稳性和微弱性造成对运动想象脑电信号的分类存在特征提取困难,分类结果不理想,分类性能受噪声影响明显等问题。针对上述问题,提出一种基于因子分析(Factor Analysis, FA)模型的噪声稳健运动脑电信号分类方法。首先利用FA模型对脑电信号中存在的噪声分量进行抑制,针对重构信号可分性较差的问题,将其转换至功率谱域,进而提取三维能够反映不同运动状态的功率谱特征,最后利用支撑向量机(Support Vector Machine, SVM)分类器对所提特征向量进行分类判决。基于Graz数据的验证实验表明,所提方法可以明显提升低信噪比条件下的分类性能,在实际工程应用中具备较强的推广泛化能力。
中文关键词:脑电信号分类  因子分析模型  特征提取  噪声稳健  
 
NOISE ROBUST EEG SIGNAL CLASSIFICATION METHODBASED ON FACTOR ANALYSIS MODEL
Abstract:In view of the difficulty of feature extraction and the obvious influence of noise on the classification of motor EEG signals, this paper presents a noise robust EEG signal classification method based on factor analysis(FA) model. Firstly, FA was utilized to suppress the noise components in EEG signals, and then the original signals with poor separability were converted to the power spectrum domain and the three-dimensional power spectrum features with good separability were extracted. Finally, Support Vector Machine (SVM) classifier was used to classify the feature vectors. The verification experiment with Graz data shows that the proposed method can significantly improve the classification performance under the condition of low SNR, and has a strong generalization ability.
keywords:Classification of EEG signals  Factor Analysis Model  Feature Extraction  Noise Robust  
查看全文   查看/发表评论   下载pdf阅读器