Research On Attention Training Of VR Flight Simulation Based On Deep Learning

With the rapid development of compute and virtual reality technology,pilot technical training through simulated aircraft has been widely adopted by major military scientific research institutions,which can effectively reduce resource overhead and training costs.Attention is the ability of people’s psychological activities to point and focus on something.To a certain extent,it can reflect the training personnel’s participation in the current training task and it’s an important aspect of performance evaluation in flight simulation training.Deep learning technology has achieved unprecedented achievements in many fields,and studies have shown that neural networks can be applied to EEG signal processing.This paper focuses on the evaluation of the attention of flight simulation training personnel,based on eye tracking and EEG technology,from two aspects: the behavioral characteristics of exogenous visual attention and the identification of endogenous attention.The specific content and innovation points are summarized as follows:(1)In the process of simulated flight training,visual attention is an important manifestation of the attention of trainers in the visual dimension,which can represent the visual search pattern and distribution law of trainers.In this paper,we design a behavioral eye-tracking experimental paradigm to study the visual attention properties of trainees driven by topdown and bottom-up attention mechanisms.By analyzing multiple eye movement indicators of the trainers under two different attention mechanisms,it is found that the trainers of the top-down attention mechanism are fast and efficient in searching for targets,and have stronger visual attention ability,while the trainers of the bottom-up attention mechanism are vulnerable to the current situation.Due to the influence of the stimulus itself,visual attention is easily distracted.Further,based on the dynamic clustering algorithm,this paper found that the regions of interest of top-down attention mechanism trainers showed different spatial distribution characteristics in the four training tasks,but the part of the flight parameters located directly in front of the simulated flight that trainers are most interested in,and the research results have important reference value for the layout of simulated flight training scenarios.(2)The representation of trainer’s intrinsic cognitive attention can more clearly reflect the trainer’s endogenous attention mechanism.This paper proposes a multi-scale convolution attention recognition model based on EEG and deep learning.The model comprehensively considers the fine-grained and computational complexity of the temporal and spatial features of EEG signals,and can be extracted through temporal and spatial convolution kernels of different sizes.Richer EEG signal features are fused to achieve complementary features at different levels.Specifically: Through statistical analysis of the questionnaire results of 12 trainers,it can be seen that there is a certain degree of correlation between task difficulty and attention.Inspired by this,this paper designs a training personnel attention-inducing experimental paradigm,and collects 12 people.EEG data during real simulated flight training.Furthermore,this paper designs a number of comparative analysis experiments and ablation experiments to explore the detection performance of the model on four self-built datasets and the influence of EEG sample length on the performance of the model.The experimental results show that the model can achieve the best results when the length of the EEG sample is 3s.On the EEG data of 12 people,the model achieves an average classification accuracy of 91.4% and macro score of 60.12%,which are better than three deep learning models and SVM algorithms.The model can effectively identify the attention of trainers,and the spatiotemporal feature extractor in the model plays a huge role.To sum up,this paper designs a new experimental paradigm and EEG data analysis method to study the visual attention behavior representation of trainers and the identification of endogenous attention levels.The two aspects of research jointly reflect the attention ability of flight simulation trainers,which can provide technical support for the comprehensive development of training personnel performance evaluation.