Research On Attention Level Training And Evaluation System Based On BCI-VR

There has been growing interest in using neural networks to recognize human brain attention levels in recent years.The analysis of electroencephalogram(EEG)signals is a promising approach for cognitive neuroscience and clinical neuroscience studies.However,most current attention training methods are relatively simple,and there is limited research on spatial cognitive training systems incorporating brain-computer interface(BCI)interaction paradigms.This study proposes a training system for attention level that combines a BCI and virtual reality(VR)to enhance the efficacy of existing attention training systems.The effectiveness of the system is evaluated by analyzing the EEG signals of subjects before and after training.Additionally,this paper explores a brain signal classification method based on SE-Inception-EEGNet in light of the evaluation results.Firstly,we use the theory of human attention to design a training system that combines a BCI under a VR environment.The system includes cognitive training and testing components,with the training module taking the form of a virtual table tennis game and the testing module a virtual whack-a-mole game.EEG signals are recorded using the BCI technology during the testing task to assess changes in attention before and after training.To extract and classify features from EEG signals,we use the Permutation Conditional Mutual Information(PCMI),XGBoost,and Random Forest in the testing task.Finally,the effectiveness of the training system is evaluated based on these results.Secondly,this study proposes a brain signal analysis model,SE-Inception-EEGNet neural network,as the second contribution.This model combines the SENet and Inception modules to enhance the feature representation ability of the EEGNet network model for different channels and spatial positions.The aim is to further improve the feature representation ability of EEG signals.The EEG signals collected during the virtual whacka-mole experiment are extracted and classified before and after training.Different neural network models’ evaluation indicators are compared to validate the SE-Inception-EEGNet network model’s superiority in EEG signal analysis.To confirm the efficiency of the evaluation system and SE-Inception-EEGNet network model,we recruited 15 participants to validate the feasibility of the attention level evaluation system.The effectiveness of the proposed SE-Inception-EEGNet network model was established by analyzing EEG data from the testing task and comparing it with other network models.