基于语谱图的管制员疲劳检测方法  

Controller Fatigue Detection Method Based on Spectrogram

作　　者：王剑辉[1] 陈超[1] 张时雨 梁海军 WANG Jian-hui;CHEN Chao;ZHANG Shi-yu;LIANG Hai-jun(Civil Aviation Flight College of China,Guanghan 618000,China)

机构地区：[1]中国民用航空飞行学院,四川广汉618000

出　　处：《航空计算技术》2023年第1期19-23,共5页Aeronautical Computing Technique

基　　金：中央高校基本科研业务费专项资金项目-学生科研项目资助(XSY2022-29);四川省科技计划项目资助(2022YFG0210);2022年中国民用航空飞行学院智慧民航专项项目资助(ZHMH2022-009)。

摘　　要：管制员作为空中交通管制的参与者,检测其工作状态对飞行的安全与正常有着重要意义。针对管制员工作过程中不同状态下的陆空通话记录,提出一种基于管制员语谱图的疲劳检测方法,通过对采集到的正常与疲劳两种状态下的陆空通话数据进行数据分割及预处理,采用傅里叶变换进行语音信号特征可视化,获得管制员语谱图;针对管制员语谱图数据搭建ConvNeXt模型,对模型进行训练学习不同状态下管制员陆空通话语谱图特征;通过实验得出所提方法对管制员疲劳状态检测准确率达97.1%,验证了方法的有效性,可以为管制员工作安排提供参考依据。As a participant of air traffic control,it is of great significance for the safety and normality of flight to detect the working state of controllers.In this paper,a fatigue detection method based on the controller speech spectrogram is proposed for the land-air call records under different states during the working process of controllers.Through data segmentation and preprocessing of the collected land-air call data under normal and fatigue states,the controller speech spectrogram is obtained by using Fourier transform to visualize the features of speech signals.Then,a ConvNeXt model was built based on the controller speech spectrum data,and the model was trained to learn the characteristics of the controllers′ speech spectrum under different states.Finally,the experiment results show that the proposed method can detect the fatigue state of controllers with 97.1% accuracy,which verifies the effectiveness of the method and provides a reference for the work arrangement of controllers.

关 键 词：管制员 陆空通话 语谱图 ConvNeXt 疲劳检测 

分 类 号：V355[航空宇航科学与技术—人机与环境工程] TP391[自动化与计算机技术—计算机应用技术]