| The human brain is an extremely complex system,and uncovering its modes of function remains an important challenge for experts in neuroscience.EEG signals is a kind of neurophysiological activity produced in a specific part of the brain,and it is also an important way to understand the brain function.Nowadays,with the development of machine learning,deep learning and other technologies,new perspectives and research methods are provided for the study of EEG signals,and further research on EEG signal is more and more profound.In modern medical diagnosis,EEG signals is a very effective way to diagnose a variety of brain diseases and mental system diseases（such as brain trauma,depression,epilepsy,anxiety,etc.）,especially in the diagnosis of epilepsy is very important.At present,the research of automatic recognition and classification of epileptic EEG signals mainly includes three aspects: feature extraction,feature selection and classification.This paper carries out research from three aspects of feature extraction,feature selection and classification.The main work is as follows:First of all,in the phase of feature extraction,the characteristics of EEG signals are studied from different aspects.Fuzzy entropy,permutation entropy and Hurst index which can accurately reflect the state of brain activity and nonlinear properties are used as features.A feature extraction method based on the combination of discrete wavelet transform and empirical mode decomposition is used to extract the features.Secondly,this paper uses m RMR algorithm to select EEG features,calculates the importance of mixed features obtained from discrete wavelet transform sub signals and empirical mode decomposition sub signals,selects the feature variables with the highest importance as the best feature subset,and establishes a random forest model to evaluate the contribution of features to verify the superiority of mixed features over independent feature extraction methods.Then,in the classification research,the classification algorithm based on robust probabilistic collaborative representation is adopted.The algorithm determines the category of the sample based on the possibility that the test sample belongs to each category,which has clear probability interpretability.Through the comparison of discrete classification algorithms such as support vector machine and linear discriminant analysis,the results show that the classification model based on DWT and EMD mixed feature extraction combined with probabilistic cooperative representation proposed in this study can fully extract the EEG features of epilepsy and accurately classify the activity of epilepsy.Compared with other EEG classification models,it is proved that the classification model has higher accuracy and sensitivity.Finally,this paper further carries out the auxiliary diagnosis of epileptic seizure activity through machine learning algorithm.Firstly,load EEG data,and then automatically normalize and preprocess EEG signals,extract time-frequency-domain features and features based on DWT and EMD.Finally,identify epileptic seizure activity under different classification scenes through LDC,KNN,SVM and R-Pro CRC classifiers,and display the results in the form of charts,Through simple operation and visual interface,it provides help for doctors and relevant personnel to study epileptic EEG signals. |
