The Research Of Brainprint Recognition In Multi-task State

Nowadays,the security of personal information becomes more important.How to carry out identity identification safely and effectively has become an important topic.The personal recognition based on EEG(Brainprint Recognition)has received more and more attention.Different from the traditional identification features which existed a wide range of defects;brainprint has the unique advantages of high concealment,cannot be stolen,cannot be imitated,and must live in areas.So far,there are many methods and applications for brainprint recognition while almost all studies have some limitations that Most of the methods of personal Identity based on EEG signals are requiring the implementation of a specific task or need to be collected in a fixed environment for analysis.There are great limitations for the development of the application of brainprint recognition with these constraints.This thesis focuses on the study of brainprint recognition in multi-task state,and studies the brainprint recognition of EEG signals collected in different environments and different tasks.This thesis studies the phase synchronization feature of EEG signal.Then functional brain network and deep belief network were used to identify the brain striations in multi-task state.This thesis had brainprint recognition based on EEG signals under multi-task state.The main work of this thesis contains following three parts:1)This thesis proposed a method that take the phase synchronization as a public feature extraction way which is different from the traditional feature extraction of the brainprint feature that use only one ways for one class of task and the phase synchronization features were used for multi-tasks brainprint recognition.At the meantime,the synchronization of EEG signal is studied and analyzed,and the mean phase locked value is proposed to measure the phase synchronization.2)This thesis proposed that using the functional brain network for the second brainprint feature extraction.In order to facilitate the personal identification,we construct the functional brain network by using symmetric matrix of mean phase locking value.The important attributes of the functional brain network: node of degree,global efficiency,clustering coefficient are combined as features for brainprint recognition.The topographic maps of brain were drawn for comparing the consistency and difference of intra-and inter – class during brainprint recognition.At the same time,it was found that the global efficiency and the clustering coefficient of the brain network always showed a positive correlation trend.3)We use the deep belief network to do the multi-task brainprint recognition.The phase information of EEG signals in different tasks and different environments is taken as the feature to train the network model and the parameters of the network are analyzed.It is found that for different dataset,the appropriate number of layers and neurons also different,which can be used for better learning for features and effective identification.In this part,we study a hybrid pattern data set containing two different tasks and the subjects in the mixed data set execute two different types of tasks.The results show that the proposed method can effectively identify the mixed data sets of different tasks.For 32 subjects,the recognition rate is exceeding 96%.In this thesis,the phase synchronization information of EEG signal is analyzed and we study the brainprint from the phase of the signal,which is different from the research studied based on of the amplitude of EEG signal.This thesis innovatively proposed that take the phase synchronization as the method of common feature extraction under multi-class tasks brainprint recognition and had experiments on five dataset.Good recognition results are achieved of 9 classes,12 classes,14 classes,20 classes and 32 classes.The accuracy rates were 99%,98%,99%,98% and 96%,respectively.These results illustrates that phase synchronization is effective as a common feature extraction method for multi-task EEG signals.This work breaks through the limitation of task and environment of brainprint recognition in some stage which makes the method of brainprint recognition more generalization.The method proposed in this thesis has some reference value for how to improve the limitation of the existing EEG identification methods mostly confined to specific task or environment.