Research On EEG Information Recognition Of Epilepsy And Mental Abnormality Based On Machine Learning

Epilepsy and schizophrenia are common brain diseases.With the improvement of living standards,people are paying more and more attention to the pathogenesis and treatment of these two diseases.Brain disease EEG(Electroencephalogram)signal processing and recognition algorithm is an important means to realize bioelectrical signal intelligent medical treatment.Doctors can combine prior knowledge and related theories to make judgments on EEG and formulate medical treatment methods for individual patients.However,this subj ective judgment method is extremely error-prone and inefficient.Advanced signal processing technology,deep neural network and other technologies provide new methods and new means for fast and precise smart medical treatment.Therefore,it is important to use advanced signal processing and machine learning technology to realize automatic detection,recognition and diagnosis of brain diseases.(1)A signal feature extraction algorithm based on wavelet bispectral energy entropy and color moment is proposed for epileptic signal processing and classification recognition.According to the characteristics of EEG signals(non-stationary,non-linear,non-Gaussian,etc.),the wavelet transform and bispectrum analysis methods are combined to give the advantages of wavelet transform and high-order spectrum estimation signal processing methods,and the wavelet bispectrum theory is obtained.In this paper,the feature vector of wavelet bispectral energy entropy and color moment is used as the method of feature extraction of epileptic EEG signals,to make classification and identification.The results show that the algorithm can effectively distinguish between ictal and interictal epileptic EEG signals and prepare for the next automatic identification work.(2)A classification algorithm of TWSVM usin g genetic algorithm is proposed.The genetic algorithm is used to optimize the twin supp ort vector machine.The penalty parameters and Gaussian kernel function parameters can be determined efficiently,then the optimal model is obtained.The results showed that the new algorithm improved the sensitivity of clinical epileptic EEG signals to 92.40%,the accuracy increased to 94.47%,and the specificity was 84.74%,and the AUC(Area Under Curve)was 96.308%.Among them,the highest accuracy of individual patients reached 99.05%.This new mechine learning method avoids the deviation of the diagnosis result due to subjective factors or the difference of judgment standards and improves the AUC by about 1%compared with the used machine learning method for unified data sources,which creates a solid foundation for accurate judgment and prevention of epilepsy in practical medicine.(3)For the EEG signals of patients with schizophrenia,an improved EEG automatic recognition algorithm based on improved VGGNet is proposed.The VGGNet network with different network layers and different convolution kernel sizes is constructed for schizophrenia EEG signals.Automatic identification is carried out.By comparison,the network model with the highest classification accuracy is selected:a 13-layer VGG network with a convolution kernel size of 5*5.Using the deep learning algorithm and the new machine learning algorithm proposed in the third and fourth chapters of this paper to respectively classify the schizophrenia EEG signals.The experimental results show that the accuracy based on the improved VGG network deep learning algorithm is 84.34%,which is better than the new machine learning algorithm,and the average accuracy is 81.45%.In conclusion,the algorithm proposed in this paper can detect the hidden state of brain diseases in time,and actively intervene in testers with epilepsy or schizophrenia tendency,so as to realize automatic recognition of EEG signals of brain diseases and complete diseases.An early assessment to achieve early prevention,early detection,and early treatment of brain diseases.The automatic detection and analysis of brain diseases can distinguish the tester's neural patterns,effectively classify data with specific types of brain diseases,manage diseases more effectively,and improve the diagnosis and treatment of diseases.