The Study On The Extraction Of Epilepsy Features In EEG Signals

Epilepsy is a common neurological disease caused by abnormal discharge of brain neurons.There are 60 million epileptic patients in the world,because of the unbalanced development,in countries or regions where medical conditions are scarce,most epileptic patients do not receive effective treatment.The most effective method of clinical diagnosis of epilepsy is electroencephalogram(EEG)when judging whether the patient suffers epilepsy.Because of the large number of EEG data collected and mixed with certain background noise,doctors have strong subjectivity in judging epilepsy by observing EEG waveform characteristics with naked eyes,which leads to low efficiency and accuracy of epilepsy diagnosis,thus delaying the treatment of epileptic patients.During the seizure of epilepsy,the EEG signal contains a lot of epileptic characteristic information.Therefore,it is very necessary to study the extraction,recognition and classification of epileptic features.The accurate extraction of epileptic features from epileptic EEG signals has great significance for the prevention and treatment of epileptic diseases.Based on the summary of the epileptic feature signal extraction research,this paper put forward two methods of feature extraction:1.This paper use the method of feature extraction of epileptic signal based on wavelet transform and sample entropy.Firstly,the brain signals are decomposed by wavelet transform,and the sub-bands with different resolutions are obtained.After wavelet decomposition,the main sub-bands contain a large number of characteristic waveforms,such as sharp wave and spike wave.Then the sample entropy of each sub-band is calculated,and the size of the entropy value is used to represent the characteristics of the characteristic waveform.2.Empirical mode decomposition(EMD)can decompose epileptic EEG into intrinsic mode functions(IMF)of different scales.The main IMF components include the information of signal fluctuation characteristics and characteristic trend.In this paper,EMD algorithm is used to decompose epilepsy signals to obtain different scales of IMF components.Then the mean and fluctuation index of the main IMF components and sample entropy are obtained.By comparing the difference between mean value,fluctuation index and sample entropy of IMF component during epileptic seizures and epileptic intervals,it is used to represent the characteristics of the characteristic waveform.After epileptic feature extraction,the extracted results are input into the extreme learning machine to further verify the effect of feature extraction.The simulation results show that the above methods are effective in extracting epileptic features.Both methods have high classification accuracy,and it is higher than other similar research.