视线追踪联合脑电的飞行员操纵意图检测方法  

Pilot Manipulation Intention Detection Methodbased on Eye Tracking Combined with EEG

作　　者：蒋光毅 王长元[2] 刘洲洲 JIANG Guangyi;WANG Changyuan;LIU Zhouzhou(School of Computer Science,Xi’an Aeronautical Institute,Xi’an 710077,China;School of Computer Science and Engineering,Xi’an Technological University,Xi’an 710021,China)

机构地区：[1]西安航空学院计算机学院,西安710077 [2]西安工业大学计算机科学与工程学院,西安710021

出　　处：《西安工业大学学报》2024年第1期23-31,共9页Journal of Xi’an Technological University

基　　金：国家自然科学基金项目(52072293)。

摘　　要：为解决基于飞行员行为与生理数据的飞行员意图分析问题,采用视线追踪设备和脑电仪实时记录飞行员视觉注视行为和脑电信号,并采用转换编码器-卷积神经网络模型(Transformer Encoder-CNN)进行操作意图的分类。通过在模拟飞行环境下开展飞行训练实验,将特定感兴趣区域的视觉注视行为联合飞行操作数据,实现不同时间段的飞行中飞行员高度、空速、姿态、航向四种操纵意图进行精确检测分类,形成操纵意图数据集,供神经网络模型进行机器学习。实验结果表明:该方法的神经网络模型对操纵意图分类识别准确率达到92%,能够有效检测飞行员的操纵意图,准确率提高了6%。The paper aims to study the method for analyzing pilots'intention based on pilot behavioral and physiological data.Line of sight tracking devices and Electroencephalogram(EEG)devices are used to record the real-time visual gaze behavior and EEG signals of pilots.And the Transformer Encoder-CNN neural network model is employed to classify operational intentions.Flight training experiments were conducted in simulated flight environments.The visual gaze behavior of specific area of interest is combined with flight operation data to achieve precise detection and classification of altitude,airspeed,attitude,and heading during flight in different time periods,forming a data set of manipulation intentions for machine learning by the neural network model.The experiments show that the neural network model of this method has an accuracy of 92%in the classification and recognition of manipulation intentions,increasing by 6%.It can be concluded that the method can effectively detect the pilot's manipulation intentions.

关 键 词：脑电 视线追踪 神经网络 操纵意图检测 

分 类 号：V328.1[航空宇航科学与技术—人机与环境工程]