Research On Recognition Of Epileptic EEG Based On Synchronization

Epileptic EEG is a typical EEG signal,and research on the activity of epileptic EEG can be used to diagnose and treat the epilepsy,which will have important significance.In this article,the synchronous characteristic of epileptic EEG is considered,and a series of good algorithms of epileptic EEG are proposed for different analysis methods of EEG.These algorithms are combined to establish a better performance algorithm of epileptic EEG based on synchronization.First of all,a better preprocessing algorithm of EEG will be selected.Currently,the preprocessing algorithms of EEG mainly include Fourier window transform(FWT),wavelet transform(WT)and empirical mode decomposition(EMD).According to the real EEG signals,these three algorithms are used to remove the noise of different EEG,and the signal noise ratio(SNR)and mean square error(MSE)are used to evaluate the results of different algorithms.The result shows that the pretreatmental effect of EMD algorithm is better.Then,the above result shows that the pretreatmental effect of EMD algorithm is better.There is a large error in the selection of intrinsic function function(IMF)component,and low recognition rate is obtained,so an empirical mode decomposition algorithm based on correlation analyze is proposed in this paper.The correlation between each IMF andthe original signal is analyzed,and the IMF with the highest correlation is selected to represent the original signal and extract the feature.Two kinds of algorithms are used to classify different signals,and six kinds of classification accuracy and ROC curve are used to evaluate the experimental results.Experimental results show that the empirical mode decomposition method based on correlation analysis has better recognition effect.Second,a feature test algorithm based on Kruskal-Waills is proposed,the most reflect the essential characteristics of EEG are selected from many features of EEG.This algorithm is a statistical test method,and the test features are valid when the test statistic P value is less than 0.005.The experimental results show that the accuracy of classification more than 85% for the effective eigenvalues,which proves that the algorithm can effectively select the EEG features with better performance.Then,a phase lock algorithm based on Hilbert-Huang transform is proposed.This algorithm can solve some problems of traditional phase lock algorithm,such as band-pass filtering loss information,selection of target frequency band is inaccurate and so on.The epileptic EEG was acquired by Indian scholar Varun Bajaj,et al and used as the classified object,and the features of EEG were extracted by the traditional phase lock algorithm and the phase lock based on Hilbert-Huang transform.The results show that the phase lock algorithm based on Hilbert-Huangtransform is easier to distinguish different types of EEG.Finally,the above research results are combined,and a new identification process of epileptic EEG is given.In this paper,the EEG will be preprocessed by EMD algorithm based on correlation analysis,and the extracted features are tested Based on the Kruskal-Waills algorithm,then,the selected effective feature index and the extracted feature by phase lock algorithm based on Hilbert-Huang transform are combined,and a complete identification algorithm of epileptic EEG based on synchronization is established.The real epileptic EEG signals are used to classify for this algorithm,and the results are compared with other algorithms with the same EEG signals.Experimental results show that recognition rates of the recognition algorithm of epileptic EEG based on synchronization both are 100 percent for between normal EEG and epileptic EEG with seizures and between epileptic EEG without seizure and epileptic EEG with seizures,and the recognition accuracy is as high as 96.25% for five kinds of epileptic EEG,and the algorithms proposed this paper has higher classification accuracy and better classification result than other algorithms.