基于脑电信号的飞行员认知负荷实时监测评估系统  

Real-time Mental Workload Monitoring and Evaluation System Based on EEG Signals of Pilots

作　　者：李葳宁 韩宗昌 邢晨光 Li Weining;Han Zongchang;Xing Chenguang(China Institute of Aeronautic System Engineering,Beijing 100012,China)

机构地区：[1]中国航空系统工程研究所,北京100012

出　　处：《航空科学技术》2024年第11期95-103,共9页Aeronautical Science & Technology

摘　　要：对飞行员在执行空战任务时的认知负荷状态进行实时监测和评估,对于保障执行任务的安全和高效具有重要作用。本文基于脑电信号(EEG)提出了动态图卷积-长短时记忆(DGCN-LSTM)网络认知负荷评估模型,该方法基于动态图卷积网络提取脑电的空间拓扑特征,并通过LSTM网络在时间维度上融合特征在不同时刻的时序信息,最终融合特征信息利用全连接层构建分类器,进行认知负荷状态的评估。为验证该算法的可行性,试验范式通过建立飞行任务仿真平台模拟多种典型的空战任务,设置复杂度不同的任务场景以诱发飞行员不同水平的认知负荷状态,采集被试者脑电信号用于模型训练与评估。在本文试验采集的样本数据集中,该算法在认知负荷分类的准确率达到89.08%,参数量为1.24M,性能优于其他基于支持向量机(SVM)等传统机器学习、卷积神经网络(CNN)、循环神经网络(RNN)、图卷积神经(GCN)骨干网络的算法模型,能够实现飞行员较准确的认知负荷实时监测评估。Real-time monitoring and assessment of pilots’cognitive load status when performing air combat missions play an important role in ensuring the safety and efficiency of missions.Based on EEG signals,this paper proposes a DGCN-LSTM cognitive load assessment model.This method extracts the spatial topological features of EEG based on dynamic graph convolution networks,and fuses the temporal information of features at different locations in the time dimension through the LSTM network.Finally,the spatial and temporal features are extracted and fused to evaluate the cognitive workload status via the fully connected layers as a classifier.In order to verify the feasibility of the algorithm,the experimental paradigm simulated a variety of typical air combat missions by establishing a flight mission simulation platform,setting up mission scenarios with different complexities to stimulate pilots’different levels of cognitive load states,and collecting the subjects’EEG signals for model training and evaluation.In the sample data set collected in this article's experiment,the average accuracy of this algorithm in cognitive load classification reaches 89.08%,with 1.24M parameters of model,and its performance is better than other algorithm networks based on traditional machine learning,CNN,RNN,and GCN backbone.This work is proved to realize real-time mental workload estimation of pilots accurately.

关 键 词：认知负荷评估 脑电信号分析 图神经网络 LSTM 实时监测系统 

分 类 号：TN99[电子电信—信号与信息处理]