Research On Physiological Mechanism,Recognition And Regulation Of Emotion Based On EEG Signal

Emotion plays an important role in cognition and behavior,such as perception,memory,learning and decision,emotion recognition and regulation has a substantial reality and broad application prospect in the fields of healthy,working lives,humanmachine interaction and so on.EEGs reflect the neural activity of our brain where the emotion generates.And because of its high temporal resolution and beyonding humun control,EEG became an important technique for emotion research.However,the existing EEG-based emotion classifiers were built under the same background,same individual and some certain time,the generalization is a huge challenge in the real application.And the lack of objective pysiological indicators to evaluate the effectiveness of regulation technique is a serious problem exsisted.To solve the above problems,the following main contents are studied:Firstly,for a further understanding the physiological mechanism,this paper firstly designs multi-time,multi-background and multi-individual emotion induced experiments respectively,with a total of 143 person-time,and EEG response to these three factors were first verified from the feature level.These results lay a theoretical foundation for building a generalized emotion classifier.Secondly,this paper proposed many methods to improve the generalization of emotion classifiers.(1)To handle the cross-time problem,we proposed learning multitime information algorithm,the average performance is significantly improved by 9.4%,and the the maximum value of 3-class classification rates achieves 86%;The similarity between resting EEG patterns in different time is proposed and proved to be an indicator of the performance of cross-time emotion classification,we then significantly improved the generalization of an emotion classifier by eliminating the resting baseline.(2)To handle the cross-background problem,IHS is proposed to separate the emotional activity and others that include those related to somatosensory processing.Then we employd support vecotor regression,combined with the cross-task feature selection,the average coefficient coefficient achieved has been improved to 0.8 using fewer features.(3)To handle the cross-subject problem,weighted ensemble learning generic information is employed and the cross-subject classification performance has been improved by 6.36%;the resting EEG patterns could predict the classification performance and improve the precision of recognition dramally by eliminating the mood baseline of each subject.Finally,this paper designs tDCS-based emotional attention bias experiment,and employs the brain laterality and ERP.The results shows that tDCS can improve the ateralization of the brain and attention bias toward the external emotional stimulus in the healthy.Moreover,in the long-term treatment of depressed patients,we find that tDCS can improve the ability of emotion regulation,reduce attention to negative emotional stimulation while improve the attention to positive emotional stimulation.In summary,this paper conducts a number of innovative researches on the emotion,and obtains many significant findings on the response mechanism,recognition and regulation of emotion based on EEG signal.