基于DWT、MEMD和模糊熵的脑电信号特征提取与分类研究

Research on feature extraction and classification of EEG signal based on DWT,MEMD and fuzzy entropy

针对脑电信号分类准确率不高导致脑控设备控制稳定性差的问题,提出一种基于离散小波变换(DWT)、多变量经验模态分解(MEMD)和模糊熵的特征提取与分类方法。首先,利用DWT将脑电信号分解成一系列窄带信号;其次,利用MEMD对子带信号进行分解,得到一系列本征模函数(IMFs),选择合适的IMFs进行信号重构,利用模糊熵算法对信号提取特征,作为实验的特征向量;最后,使用支持向量机(SVM)进行分类。利用脑机接口(BCI)大赛数据作为验证集,验证了该算法的有效性,使分类精度提高到了96.2%,同时解决了经验模态分解(EMD)中频带覆盖较广的问题。

Aiming at the problem that the low classification accuracy of EEG signals results in poor control stability of brain-controlled equipment, a feature extraction method was proposed based on discrete wavelet transform(DWT),multivariate empirical mode decomposition(MEMD)and fuzzy entropy.Firstly, the electroencephalogram(EEG)signal was decomposed into a series of narrow band signals with DWT,and then the sub-band signals was decomposed with MEMD to get a set of intrinsic mode functions, which were called intrinsic mode functions(IMFs).Secondly, the appropriate IMFs for signal reconstruction were selected.Then, the fuzzy entropy algorithm was used to extract features from the signal as an experimental feature vector.Finally, support vector machines(SVM)were used for classification.By using the BCI competition data as a verification set, the effectiveness of the algorithm was verified, and the problem of wider EMD mid-band coverage was solved.