The Identification And Classification Of Epileptic EEG Signal Based On ITD Method

With the rapid development of modern information technology and biological neuroscience,related research on a series of electrophysiological signals such as electroencephalogram(EEG)and electrocardiography(ECG)has gradually attracted high attention in the medical field and machine learning field.EEG signals are the most important electrophysiology in the body.EEG signals contain a large number of physiological and pathological information,and play a crucial role in the diagnosis of human brain diseases such as epilepsy,depression,Alzheimer's disease,etc.On the other hand,the acquisition of EEG signals is convenient,non-invasive,and costly.Therefore,the research on EEG signals has attracted more and more attention.The main work of this article on EEG signals is as follows.On the one hand,a denoising model based on total variation regularization and non-sampling wavelet transform is proposed.The removal of noise is the first step in the identifica-tion and analysis of EEG signals.Total variation regularization is generally applied in the field of image processing.More extensive,it can protect the edge information of the image while removing the noise.The wavelet transform has the characteristics of multi-resolution analysis,with time-frequency locality,so it is very suitable for analyzing non-linear non-stationary sig-nals such as EEG signals.However,the application of wavelet transform denoising can easily cause local oscillation at the edge of the signal.In this thesis,we combine the total variation regularization with wavelet decomposition to give a new denoising model and apply the split bregman iterative algorithm to solve the model.In the experimental part,compared with the ex-isting methods,this method has obvious advantages,and its application to EEG signal denoising can significantly improve the signal-to-noise ratio.On the other hand,a new set of features is extracted for the classification of epileptic EEG signals.Based on the nonstationarity and non-linearity of EEG signals,this paper applies the intrinsic time-scale method to the time-frequency analysis of EEG signals.Overcome the end effect of Hilbert-Huang transformation.Based on the waveform characteristics of epileptic EEG signals and normal EEG signal signals,this paper proposes to use the cascade of the mean,vari-ance,kurtosis,and skewness of the instantaneous and instantaneous amplitude characteristics of epileptic EEG signals and normal EEG signals as the characteristics of EEG signals,and to prove the extracted features through experiments.It is feasible;In the end of this study,the features of EEG signals are classified by two classifications and three classifications using the support vector machine classifier,and compared with the classification results of the existing literatures,it is concluded that the features selected in this paper have obvious superiority.