| In the railway field train driving fatigue is one of the important factors leading to accidents.As the most critical safety role among railway transportation participants,train drivers are extremely prone to fatigue due to their perennial working conditions of harsh working environment,long task time,heavy load and irregular work and rest.At present,with the further improvement of the automation level of the railway industry,drivers shift from a traditional physical workload to a high cerebral cognitive workload,and the continuous vigilant work will make the driver’s brain fatigue phenomenon more prominent,which seriously threatens the safe operation of trains.This paper takes the long time monotonous vigilance task and operation behavior characteristics of train driver in-transit driving process as the entry point to explore a method of detecting train driver in-transit fatigue state based on non-contact millimeter wave radar physiological data and driving behavior data,contains 51 figures,43tables and 95 references,the specific main work is as follows:(1)The theoretical model mechanism of FMCW millimeter wave radar physiological sign detection was explored.Based on the in-depth analysis of the performance parameters and applicability of different types of radars,a radar non-contact physiological data acquisition platform was established based on AWR1642 millimeter wave radar and DCA1000 acquisition board,and a method of driver on-the-road fatigue state detection based on non-contact physiological characteristics is proposed.(2)Based on the phase-mean phase extinction and multi-chrip incoherent accumulation for human target location,a band-pass filtered primary signal separation and optimized modal decomposition secondary signal separation reconstruction algorithm was proposed to realize the FMCW millimeter wave radar non-contact physiological sign data recognition.Through accurate comparison and evaluation experiments with medical portable Holter dynamic ECG instrument,it was found that the overall average error of heart rate based on FMCW recognition was 1.08 bpm,and the total average error percentage was only 1.82%,which fully satisfies the requirements of recognition of respiratory and cardiac physiological indices related to driving fatigue.(3)Based on an in-depth study of the fatigue-inducing paradigm,a method based on the AX-CPT paradigm of long-time cognitive tasks to simulate train driving fatigue induction was proposed,and the relationship between fatigue level and physiological indicators and overall performance during long-time cognitive tasks was experimentally studied,and the driving fatigue cut-off point was reasonably determined from the perspective of multi-dimensional data,which provides a scientific basis for the effective calibration of the fatigue state of subsequent physiological data and the data input of fatigue detection models.(4)The differential changes of physiological indicators before and after fatigue were analyzed in depth from the time domain of respiration,heartbeat,heart rate variability and frequency domain of heart rate variability characteristics,and a method of calculating the characteristic parameters of non-contact physiological indicators of millimeter wave radar and a fatigue detection model with multimodal data fusion were proposed,and the detection accuracy of the model was verified and analyzed,and the proposed SVM model obtained a better The proposed SVM model obtained better prediction results compared with other machine learning models,and the highest fatigue classification prediction accuracy reached 82.8%.The fatigue level assessment model can be constructed to further subdivide the fatigue level based on the number of fatigue features on the basis of the fatigue classification model,which makes up for the deficiency that fatigue cannot be predicted in a hierarchical manner. |
PDF Full Text Request