The Research And Application Of EEG Recognition Based On EEG Phase Synchonization

EEG signal is a basic physiological signal,it contains a wealth of information and can effectively record the brain's nervous system activity.At present,EEG recognition are mostly based on amplitude characteristics of EEG signals,but it ignores the phase synchronization of multi-channels EEG.EEG synchronization can reflect the brain's neural activity and can be used to speculate people's cognitive activities.The phase synchronization feature of the EEG signal can detect the weak interrelationship of different signal pairs only by the instantaneous phase relation between the signals because it ignores the influence of the amplitude.It can effectively distinguish the different consciousness tasks of the human and reflect people's recognition function.In this thesis,we mainly studied the research and application of EEG recognition based on EEG phase synchronization,emotion recognition and identification recognition.The phase synchronization theory and its concrete calculation method were studied.It studied from PS of signal to PS of EEG.With the help of Hilbert Transform,the EEG is decomposed into instantaneous amplitude and instantaneous phase.Then the 1:1 phase synchronization is used to measure the PS level of two-channel EEG signals.An emotion recognition method based on EEG phase synchronization was proposed.We take experiment on the preprocessing data of DEAP standard database and utilize the brain topographic map to display the phase synchronization features of the same kind of emotion.It is found that the phase synchronization feature of the emotion data can reflect the neural activity of different brain regions,and can be used for emotion classification.Then we utilized the principal component analysis and the sparse expression classifier to reduce the dimension and classify the emotion.The recognition rate of the second,third and fourth classes of emotion are 94.5%,87.61%and 67.04%,respectively.An identification recognition method which combined EEG phase synchronization and multiple manifold learning was proposed,this method take the vector correlation coefficient as distance between every two samples.It takes PS feature extraction from the data collected from the laboratory of the author(collect the subjects' EEG signals during the neutral video interspersed with advertisements).Then the multi-manifold algorithm and the nearest neighbor classifier based on vector correlation coefficient are used to space embedding and classify the identities of people.The average best recognition rate of the 20 subjects in the four frequency bands is 66.8%,86.4%,91.2%and 92.5%,respectively.The applications of emotion recognition and identification recognition show that the phase synchronization feature of EEG can be effectively applied to EEG tasks,and the phase synchronization feature of EEG can effectively reflect the cognitive function of brain.Meanwhile,the correlation coefficient of vector is also an effective sample similarity measurement method.