Recognition Of The Fatigue Status Of Pilots Based On Electroencephalogram Signals

As the air transport plays an increasingly prominent role in transportation,the aircrafts carry an important mission to protect people's lives and properties.Thus aviation safety has become an eternal theme and the firm pursuit of civil aviation.At present,the ensurance of aviation safety has also become a top priority for the sustainable development of civil aviation.The major factors that will affect aviation safety are mainly people,environment and machinery.In recent years,with the advancement in the design,manufacture,maintenance,etc.,the aviation aircraft developed very quickly.The reliability of the aircraft and the safety of the flight have also been significantly improved.However,the proportion of flying accidents caused by human factors has not declined,especially in terms of pilots' fatigue.Therefore,the quantitative identification of the fatigue status of pilots timely warning,and the reduction of the accidents caused by pilots' fatigue status are of great significance to aviation safety,and have become a major practical problem that needs to be solved urgently in the current practice of aviation safety management.As the development of the portable electroencephalogram signals acquisition equipment,the difficulty in acquisiting electroencephalogram signals during the fatigue test has been largely reduced.The electroencephalogram signal is named as Gold Standard of recognition of the fatigue status.This paper analyzes electroencephalogram signals,obtained during the flight simulation experiments in the CRJ-200 simulation cockpit and the physiological condition in operating the flight simulation,and puts forward the deep learning network model based on the electroencephalogram signals to identify pilots' fatigue status.First of all,we designed the simulation experiment based on the causes of the pilot fatigue,and divided the pilot fatigue into non-fatigue,mild-fatigue and extreme-fatigue according to the two-stage fatigue criterion.Then,to obtain a more accurate data of pilots' fatigue status,the collected electroencephalogram signals were analyzed accurately using wavelet packet transform to extract the four rhythms.And obtain a more accurate data of pilots' fatigue status.Finally,we proposed a deep learning network model based on electroencephalogram signals processing to identify the pilot fatigue state.Through the training and testing of the proposed model using the samples of the electroencephalogram signals obtained from the simulation experiment,it is found that the network model based on the deep sparse aotuencoding network and deep contractive aotuencoding network could better identify the pilots' fatigue status.In addition,the proposed deep learning network model can provide a new theoretical basis for pilots' fatigue status detection method.And it also provides provide a direction for the monitoring and warning of the fatigue status of pilots during the execution of the flight mission through the electroencephalogram signal in the future.