EEG Signal Analysis And Application Research Based On Nonlinear Topological Dynamic Method

With the rapid increase of people’s workload,the high-pressure working state causes frequent physical and physiological diseases,which makes people pay more attention to their mental state.The human physiological signals are closely related to thier physical activity and behavior.EEG signals are an important data source for the study of human activity and behavior.Through the study of brain signals,various diseases and mental states can be detected.The brain is considered to be a nonlinear dynamic system.Traditional EEG signal processing methods are mostly based on linear system theory,which will inevitably lose the information carried by the original signal.However,the measurement method used by common nonlinear analysis theories do not have a clear metric.In this paper,a nonlinear topology dynamic method based on Persistent Homology is used to study the analysis and application of EEG signals.The main work is as follows:1)First,the processing flow of the nonlinear topology dynamic analysis method is introduced,and its basic theory is explained.The nonlinear topology dynamic method mainly includes two aspects: the spatial embedding of the original EEG signal and the topology dynamic analysis of the point cloud.For the first aspect,this paper studies three different point cloud matrix construction methods: delay embedding,original embedding and distance matrix of EEG signals.For the second aspect,topological dynamic analysis,the process of complex filtering and topological feature calculation is studied.2)Research on classification tasks based on the mental arithmetic dataset on MIT Physionet.This paper studies the application of three point cloud construction methods in multi-channel EEG signals.The matrix constructed by these three methods is used to extract topological dynamic features for mental arithmetic state,and and perform classification experiments on mental arithmetic state and mental arithmetic performance.Finally,the experimental results are analyzed.3)Based on the nonlinear topology dynamic method,the emotional EEG signal processing framework TEEGNDA is proposed.This paper sets up experiments related to rhythms and channels,performs classification experiments on two EEG emotion datasets,DEAP and DREAMER,analyzes the results and compares them with representative literature.According to the experimental results: on the mental arithmetic task,the classification accuracy of the three construction methods is above 98%,and the cross-subject results are better than deep learning;on the emotion recognition task,the classification accuracy of the two emotion EEG datasets both exceeded 99%.Therefore,the nonlinear topology dynamic method is a powerful tool to explain the dynamics of the nervous system.It has strong interpretability for multi-channel and single-channel EEG,and has shown excellent recognition ability in different classification and recognition tasks.