Epileptic Electroencephalogram Recognition Based On Common Spatial Model Algorithm And Support Vector Machine Double Classification

Epilepsy is a chronic brain dysfunction syndrome caused by multiple etiologies and it is the second most common brain disorder after cerebrovascular disease.Seizures are directly caused by intermittent central nervous system dysfunction due to repeated sudden over-discharge of neurons in the brain.Clinical manifestations include sudden loss of consciousness,generalized convulsions,and mental abnormalities.For children with epilepsy,physical and intellectual development can be greatly affected,and more patients with epilepsy suffer physically and psychologicallyElectroencephalography(EEG)is an important tool to study the characteristics of seizures.It is a non-invasive physical examination method.Important information thatother methods of human physiology cannot provide can often be found in the EEG Monitoring the brain for abnormal firing phenomena is a major part of EEG analysis,such as the detection of spike waves,multi-spike waves and spike/slow complex waves.At present,EEG as the basis for physicians to diagnose epilepsy diseases,its diagnosis method mainly relies on manual interpretation and analysis by professional physicians,which is time-consuming,subjective,and the judgment results of different physicians for the same record may be different,so the current epilepsy detection results have a strong dependence on physicians’ experience,and there is a high rate of misdiagnosis in the diagnosis process.The rapid development of computer technology and machine learning provides a solution for the automatic detection of epileptic EEG.Whether automatic recognition technology can achieve timely,accurate diagnosis and prediction of epileptic EEG is an important part of current research.In order to realize the automatic recognition of epileptic EEG signals in various periods with high efficiency and accuracy,this paper proposed an automatic recognition technique of epileptic EEG signal based on the spatial characteristics extracted by the CSP and the dual classification mode of SVM.The main principle of this technique is to extract features of the EEG signal,including the standard deviation in the time domain,the wavelet packet energy of the time-frequency analysis,the entropy features and the spatial characteristics.Then,the above features are composed into a feature vector,so that the information in the EEG signal can complement each other.Finally,the feature vectors are sent to the SVM classifier for double classification,and the classification results of normal,interictal and seizure periods are obtained.The main innovations in this study are as follows(1)A combination method of multiple features is proposed.The features of each domain complement each other to describe the feature information embedded in EEG signals more and more comprehensively,so as to improve the recognition rate of EEG signals in each period.(2)The spatial characteristics based on the CSP compensates for the information description of epileptic EEG signals in the spatial domain.The main step of the CSP is to perform spatial filtering,which is well suited for processing multidimensional EEG signals,and it can simultaneously exploit the spatial correlation to effectively extract the spatial characteristics of EEG signals on different channels.The experimental results showed that the introduction of the spatial characteristics greatly improved the recognition rate of the three periods.(3)To improve the efficiency of recognition of EEG signals in three periods and to meet the real-time requirements in medical applications,a new double classification model based on SVM is proposed in the pattern recognition stage.The double classification model is based on the fact that the actual seizure process is always a continuous process from the normal to the interictal and then to the seizure.The double classification requires only two comparisons of epileptic EEG signals,which can effectively improve the recognition efficiency compared with the traditional direct three classification method.The database comes from the Epilepsy Research Center of the University of Bonn,Germany.It contains five types of EEG signal samples:awake with eyes open in healthy subjects,awake with eyes closed in healthy subjects,the data collected inside and outside of the epileptic foci from patients during their interictal period,and the data from patients during their ictal period.The experiments verified that the combination of spatial characteristics and other domain features could provide a more comprehensive characterization,and made significant contributions in the later classification recognition,achieving a high classification recognition rate.The double classification model based on SVM achieves higher recognition efficiency compared to other classification methods in the literature,and provides better aid in practical clinical medical testing.