Feature Extraction And Recognition Of Electroencephalogram Based On Ensemble Empirical Mode Decomposition And Random Forest

It is of great significance to automatic detection and recognition for long-term monitoring of epileptic electroencephalogram(EEG)in clinical medicine.Epilepsy EEG signals contains abundant physiological information and pathological information,which can not only provide a basis for clinical diagnosis,but also provide effective aids for the treatment of epilepsy diseases.Epilepsy EEG signals must be decomposed to obtain the useful information from them.Epileptic EEG signals are nonlinear and non-stationary like other EEG signals.Traditional methods are only suitable for linear and stationary signals,and they are not suitable for epileptic EEG signals.Therefore,we need new methods to deal with epileptic EEG signals.This article uses the Hilbert-Huang Transform to decompose epileptic EEG signals because its certain degree of adaptability.This paper proposes a new method for feature extraction and recognition of epileptic EEG based on Ensemble Empirical Mode Decomposition(EEMD)and Random Forest(RF).200 single-channel signals of interictal and ictal epileptic EEG signals were selected from Bonn data sets,a total of 819,400 points were used as samples.Firstly the EEG signal is decomposed into several Intrinsic Mode Functions(IMF)and one trend using EEMD by MATLAB.And calculate the correlation coefficient of each IMF and the epileptic EEG signals to select the appropriate IMF component.Then the effective features are calculated by SAS.The Random Forest(RF)and Least Squares Support Vector Machine(LSSVM)are used to classify the features of epileptic EEG signals by R,and the classification results of the two methods were compared.Finally we choose a better method.The experimental results shows that the classification method combining EEMD with RF is better for the classification of epileptic EEG signals during interictal and ictal.The accuracy rate of 99.60% is achieved,which is higher than the 98.00%accuracy of the LSSVM classification results.Both methods can effectively distinguish between the interictal and ictal epileptic EEG signals.The accuracy of the combination of EEMD and RF method is higher.