Research On Affective Brain-computer Interface And Emotion-related Brain Activity

Emotion plays a significant role in people’s daily lives.The maintenance of a suitable emotion can effectively safeguard one’s physical and mental health.The affective brain-computer interface(a BCI)system aims to identify people’s emotional states via brain activity signals and further attempts to regulate their emotional states,which is the realization of emotion artificial intelligence.In addition,through investigating the trajectory of brain activity and the underlying neural mechanism behind diverse emotional states,we may better understand the generation of emotional states and design more effectively a BCI system.The main contribution of this study is the development of a BCI system to detect user’s emotional states in real time,assisting emotion regulation through neurofeedback training,and exploring brain functional mechanism underlying emotional states by evaluating various types of brain signals.Firstly,we proposed an a BCI system based on electroencephalography(EEG)signals for neurofeedback training.The system could detect subjects’ emotional states in real time and deliver feedback to assist them change appropriate strategies for eliciting the corresponding feelings and achieving emotion regulation.Twenty subjects in the experimental group and twenty subjects in the control group participated in our neurofeedback training,which consists of ten experimental sessions.The experimental results revealed that,as compared to the control group,subjects in the experimental group were able to elicit the appropriate emotions more accurately following neurofeedback training,indicating an improvement in their emotion regulation abilities and demonstrating the effectiveness of our system.Moreover,we obtained brain activity patterns for each emotional state by examining the band-power-based topological maps in each experimental session.Furthermore,these patterns were steadily enhanced as the training proceeded,which may provide new insights into the underlying brain mechanism of emotion regulation.Secondly,to address the inadequacies of the above-mentioned a BCI system that takes a long time to collect training datasets,we further proposed a domain-fusion-based multi-source style transfer mapping(DF-MS-STM)transfer learning(TL)algorithm for cross-subject emotion recognition.This algorithm-based a BCI system can directly predict the emotional states of new users without the need for extra calibration runs.16 subjects participated in our online experiments,and the experimental results demonstrated that when compared to previous transfer learning algorithms,our suggested DF-MS-STM method achieves superior cross-subject emotion classification accuracy.The system has a tremendous potential for application in daily self-emotion monitoring.In addition,these subjects were divided into two groups based on the similarity of their EEG features.Substantial variations were discovered in the emotion-based brain activity patterns of the subjects in the two groups.These emotion patterns might help us understand individual differences in emotion and provide new analytical approaches for cross-subject emotion identification.Next,we used functional magnetic response imaging(f MRI)to further explore the characteristics of brain activity in different emotional states.We took advantage of the high spatial resolution of f MRI images investigate the influence of neurofeedback training based on the above-mentioned a BCI system on subject’s brain activity.Twenty subjects underwent an f MRI experiment before and after neurofeedback training.The comparison of the two f MRI experimental results revealed that,after neurofeedback training,more brain regions related to emotional brain activity could utilized to discriminate different emotional states.These results indicated that the subjects could mobilize more brain regions to elicit their emotions after neurofeedback training,indicating that EEG-based neurofeedback training might successfully enhance emotion regulation ability.Lastly,we explored an invasive techniques-based brain-computer interface and its applications in brain activity research.First,we validated the feasibility of developing a brain-computer interface system based on stereoelectroencephalography(SEEG)signals in 13 subjects who had SEEG electrodes implanted,and then analyzed the brain activity mechanism underlying the experimental paradigm of this system.The experimental results demonstrated the utility of SEEG data in brain function research.Following that,we conducted an emotional experiment on 8 subjects who had SEEG electrodes implanted.The experimental results revealed that several brain regions played important roles in emotion recognition,and emotion-related cognitive activities necessitate the collaborative activity of different brain regions,reflecting the complexity of emotional activities.In the future,these brain regions might play critical roles in the development of a BCI systems for emotion detection or closed-loop emotion regulation.