摘要
为提高运动想象脑电信号特征的区分性,提出一种基于多元变分模态分解(MVMD)的多域特征结合脑电特征提取方法。首先利用MVMD对原始脑电多通道数据进行自适应分解,然后从分解得到的固有模态函数(IMF)分量提取信号的时域特征以及非线性动力学特征,同时将IMF分量合并构造新的信号矩阵,并采用共空间模式(CSP)法对重构信号提取空间特征,进行时域、非线性动力学以及空域特征的结合,最后通过支持向量机(SVM)对此特征集分类。所提方法在BCI Competition II Dataset III数据集上达到了89.64%的分类准确率,与现有的方法比较,验证了所提方法的有效性。
To improve the discriminability of the features of motor imagery electroencephalography(EEG)signals,a multi-domain feature extraction method of EEG based on multivariate variational mode decomposition(MVMD)is proposed.Firstly,MVMD is used to adaptively decompose the original EEG multi-channel data.Secondly,extract the time domain characteristics and non-linear dynamic characteristics from the obtained intrinsic mode function(IMF)components,simultaneously the IMF components are constructed as a new signal matrix,and the common spatial pattern(CSP)approach is employed to extract spatial features of the reconstructed signals,and then we combine features with time domain,nonlinear dynamics,and space domain.Finally,this feature set is classified by support vector machine(SVM).The proposed method has achieved a classification accuracy of 89.64%on the BCI Competition II Dataset III,and the effectiveness of the proposed method has been demonstrated by comparing with the existing methods.
作者
孟明
闫冉
高云园
佘青山
MENG Ming;YAN Ran;GAO Yunyuan;SHE Qingshan(School of Automation,Hangzhou Dianzi University,Hangzhou 310018,China)
出处
《传感技术学报》
CAS
CSCD
北大核心
2020年第6期853-860,共8页
Chinese Journal of Sensors and Actuators
基金
浙江省自然科学基金项目(LY18F030009)
国家自然科学基金项目(61971168,61871427)。
关键词
脑机接口
运动想象
多元变分模态分解
特征提取
Brain-Computer Interface(BCI)
motor imagery
Multivariate Variational Mode Decomposition(MVMD)
feature extraction
