An Epileptic Electroencephalogram Signal Automatic Detection Method Based On Modified Variational Mode Decomposition

Epilepsy is a common disease of the brain.The physiological signal is a crucial gist for clinical diagnosis.Through electroencephalogram(EEG)can non-intrusive to epilepsy seizures monitoring and positioning in the human brain to a region,so the accurate and efficient through the EEG signals automatically diagnose epilepsy in computer technology has been becoming the current hot spot.In order to obtain high efficiency and stable epileptic EEG signal automatic classification method,this article obtains from the epileptic EEG signal processing algorithms,researches Variational Mode Decomposition(VMD)which is a time-frequency domain analysis algorithm,explores the improved Variational Mode Decomposition method of automatically detecting the epileptic EEG signal,and proposes two new signal processing algorithm,designes the three kinds of epileptic EEG signal model of the automatic classification algorithm.(1)With regard to the weak characterization capability and poor performance of traditional methods of single-domain EEG signal feature extraction,this paper puts forward an automatic detection of epileptic EEG signals in complex domain,which combined with VMD and Refined Composite Multiscale Dispersion Entropy(RCMDE),Refined Composite Multiscale Fuzzy Entropy(RCMDE).For a public EEG data set,the final experimental consequence demonstrates that the accuracy is 94.24%,the sensitivity is 95.58%,and the specificity is 90.64%.Furthermore,in order to resolve the number of VMD’s modes,proposes a way to judge the center frequency of the variational mode function based on the variation curve.In order to resolve the characterization ability of characteristic RCMDE and RCMFE in epileptic EEG signals,proposes a way to estimate the difference of the two characteristic box graphs under different variational mode functions.(2)With regard to VMD by ridge regression to construct constraint equations that bring about the low degree of shrinkage and estimate is not accurate,this paper uses Elastic net to improve VMD,proposes a new signal analysis algorithm named Elasticity Variational Mode Decomposition(EVMD),and proposes based on EVMD and RCMDE composite domain method to automatical detect the epileptic EEG.For a public EEG data set,the final experimental consequence demonstrates that the accuracy,sensitivity and specificity can achieve 92.54%,93.22% and 91.86%respectively,which are higher than the VMD.It attestes that the proposed EVMD algorithm has more powerful characterization ability compared with the VMD in the epileptic EEG signal processing.In addition,the proposed EVMD algorithm is tested by using simulation signals,and it is proved that it has good noise robustness.(3)With regard to the proposed EVMD algorithm like VMD is single channel signal processing method,when dealing with multi-channel EEG signals,has low efficiency and poor coordination problems,this paper generals extension will EVMD in multiple domains,puts forward a new deal directly with the algorithm of multiple signals named as multiple Elastic Variational Mode Decomposition algorithm(Multivariate Elastic Variational Mode Decomposition,MEVMD),and thus is proposed based on MEVMD with RCMDE composite domain automatic method to detect epileptic EEG signal.For public EEG data sets,the final experimental consequence demonstrates that the accuracy is 93.64%,and the sensitivity is 93.93%,and the specificity is 92.92%,compared with EVMD better classification effect not only,and fifty percent of cross validation experiment is ten times and lower,more stable,it has better generalization application potential,but compared with EVMD run longer and costs more.In addition,the proposed MEVMD algorithm is tested by using simulation signals,and it is proved that it has good modal alignment characteristics.