摘要
提出一种基于分数阶傅里叶变换(Fractional Fourier Transform,FRFT)和相关向量机(Relevance Vector Machine,RVM)的运动想象脑电信号分类方法。利用不同阶次的FRFT将脑电信号转换至分数域,在分数域提取44维分数阶特征,充分扩展特征域的同时尽可能多地从不同维度提取信号中的有用信息。利用RVM分类器进行特征选择和分类识别,在自动确定最优分类特征的同时获得理想的分类结果。基于国际BCI竞赛2003中Graz数据的实验结果表明,该方法可以获得97.51%的正确识别率,并且具有较强的泛化能力和噪声稳健性。
This paper proposes a classification method of motor imaginary EEG signals based on fractional Fourier transform(FRFT)and relevance vector machine(RVM).Different levels of FRFT were used to convert the EEG signals to the fractional domain,and the 44-dimensional fractional order features were extracted in the fractional domain,fully expanding the characteristic domain while extracting as much useful information from different dimensions as possible.The RVM classifier was used for feature selection and classification recognition,and the ideal classification results could be obtained while the optimal classification feature was automatically determined.Experimental results based on Graz data of BCI in 2003 show that our method can achieve 97.51%correct recognition rate,and has strong generalization ability and noise robustness.
作者
詹宏锋
Zhan Hongfeng(Guangdong Polytechnic of Science and Technology,Guangzhou 510640,Guangdong,China)
出处
《计算机应用与软件》
北大核心
2020年第11期146-153,共8页
Computer Applications and Software
关键词
脑电信号分类
特征提取
分数阶傅里叶变换
相关向量机
Classification of EEG signals
Feature extraction
Fractional Fourier transform
Relevance vector machine
