Research On EEG-based Emotion And Attention Recognition And Its Application

Brain-computer interface(BCI)provide non-muscular communication and control by directly translating brain activities recorded from the scalp into computer control signals and thus enable users to convey their intent to the external world.Research on BCI technology has important theoretical significance and broad application prospects.An important issue in the BCI research is how to improve system performance and develop more practical BCI system.This dissertation focuses on emotion recognition and attention detection based on BCI,and explores its potential applications.In order to explore how the human brain processes positive and negative emotions,and develop emotional interactive brain-computer interface systems,this study proposed an EEG-based BCI system to distinguish video-induced positive and negative emotions.Ten healthy subjects participated in the experiment and achieved a high average online accuracy of 91.5%.The experimental results demonstrated that the subjects emotions had been sufficiently evoked and efficiently recognized by our system.The results of offline analysis show that there are differences between positive and negative emotional patterns,especially in the forehead area of Θ band,the temporal lobe of β and γ bands.Next,we applied our emotion recognition BCI system to patients with DOC.Eight DOC patients participated in our experiment,and three of them achieved significant online accuracy.The experimental results show that the proposed BCI system could be a promising tool to detect the emotional states of patients with DOC.Secondly,in order to enhance the system performance of the above-mentioned BCI system and to study more different types of emotional brain patterns,a BCI system for fear emotion recognition was designed.Unlike positive and negative emotions,fear emotion can induce stronger emotional responses more quickly,and thus may be more suitable for clinical applications.In particular,we integrated functional brain network analysis methods and traditional spectrum analysis methods to distinguish fear emotion and neutral emotion.The two methods analyze the process of emotional processing from different perspectives,but the conclusions are consistent.Next,we studied the spectrum characteristics of the resting state EEG of tinnitus patients and compared the attention differences between tinnitus patients and healthy subjects.A total of 16 tinnitus patients and 16 healthy subjects were recruited to participate in the experiment.The results of resting state EEG experiments founded that the spectrum power value of tinnitus patients was higher than that of healthy subjects in all concerned frequency bands.And the significant difference areas were mainly concentrated in the right temporal lobe of the θ and α band,and the temporal lobe,parietal lobe and forehead area of β and γ bands.The results of task state experiments showed that the classification accuracy of tinnitus patients was significantly lower than that of healthy subjects.The experimental results indicate that tinnitus may cause the decrease of patients’ attention.Finally,this study developed a high-performance attention recognition BCI system that requires no training and is more portable.Specifically,we collected a large number of participants’ attention data using a Neuroscan amplifier and self-designed headband.Based on these attention EEG data,we established a cross-subject deep learning model and achieved high recognition accuracy.Furthermore,we deployed the deep learning model on our headband BCI system using transfer learning and fine-tuning method.The experimental results show that the transfer learning based model could achieve satisfactory attention detection results on headband device.We also designed attention regulation BCI system based on our headband,the system could output the users’ attention level in real-time.The test results also verified that our attention BCI system could detect the user’s attention level correctly and provide online feedback.