Research On Fatigue Degree Of High-Speed Rail Dispatchers Based On Facial Features

The fatigue work of high-speed railway dispatchers is a serious railway safety problem.If the high-speed rail dispatcher is found to be overly fatigued during the operation,and a warning or other measures are issued,it can avoid the railway traffic safety accident caused by fatigue.Accurately predicting the fatigue degree of high-speed railway dispatchers is a key technical issue to improve the work efficiency of dispatchers and ensure the safety of train operation.Aiming at this problem,a classification prediction method for the fatigue degree of high-speed rail dispatchers based on facial features is proposed.By recording the facial features and other information of the dispatchers during the operation of the high-speed rail,and clustering the data at the input end of the model based on K-means clustering,the optimal grading number of dispatchers fatigue based on facial feature information is calculated.On this basis,the fusion algorithm is used to calculate the fatigue grading threshold and used as the output terminal.Facial feature information(gaze time,average pupil size,blink frequency,blink duration,yawn frequency)and working time are used as input terminals,respectively based on BP(Back Propagation)pattern recognition neural network model and hidden Markov model(HMM)to establish a high-speed rail train dispatcher fatigue prediction model.According to the facial feature data of 12 high-speed rail dispatchers simulating the dispatch task,the model is trial-calculated.The research results show that the best classification number of fatigue degree based on facial feature information of high-speed rail traffic dispatchers is 3.The HMM model has a higher judgment accuracy rate in the period feature data set than in the time feature data set,and it performs better than the BP neural network in the judgment of the I-level fatigue.The BP neural network has similar judgment effects on the two data sets and is better than the HMM model,and has a more accurate prediction and judgment on the II and III fatigue.The two models can complement each other in the use of time-period feature data sets;under the time-of-day feature data sets,the BP neural network model is the best.