Research On Fatigue Driving Warning Technology For Train Drivers

Railway transportation is an important hub of transportation industry,but its accidents will pose a great threat to life safety and property safety.Fatigue driving is one of the causes of accidents.This thesis extracts and analyzes the performance parameters of audio and video performance of railway train drivers,and studies the main characteristics of fatigue driving of railway train drivers by combining with the frontier technologies such as multi information fusion.The online detection system and digital analysis judgment model are designed to realize the on-line detection and early warning of fatigue driving.The specific research contents are as follows:Firstly,the accuracy and practicability of common driver fatigue detection methods are analyzed,and the processing scheme of driver fatigue voice signal is given.The voice sample data is collected and preprocessed.This thesis studies the extraction method of speech fatigue feature parameters.In this thesis,four features are selected to detect speech fatigue,which are the mean of fundamental frequency,short-time energy,short-time average zero crossing rate and Mel cepstrum coefficient.By studying machine learning algorithm,an improved gray wolf optimization algorithm(gjgwo)is proposed.A speech multi feature gjgwo-svc classification and detection model is established.Based on MATLAB platform,the performance of the model is tested,and the accuracy of the speech multi feature gjgwo-svc classification and detection model is analyzed and verified.Secondly,the processing scheme of driving fatigue video signal is given.The fatigue feature parameters of eyes and face are extracted from the video sample data.The Gaussian skin color model is studied,and the skin color region is selected.At the same time,the face region is detected by Ada Boost classifier and Haar feature algorithm.In this thesis,the image feature method is used to locate the human eyes and recognize the opening and closing degree.According to the law of three Court and five eyes,the mouth location and opening and closing degree detection were studied.In this thesis,eye features such as percentage of eye closure time,average speed of eye closure,maximum continuous time of eye closure,average degree of eye opening and blinking frequency,and yawn features are extracted for fatigue detection of video signals.The eye multi feature gjgwo-svc classification detection model is established,and the lower video samples under different fatigue states are collected to analyze and verify the accuracy of the model.Then,by studying the principle and scheme of information fusion,a method based on decision level fusion and feature level fusion is studied,and a multi information fusion detection scheme of driving fatigue voice and video is given.According to the probability theory of Bayesian network,the dynamic Bayesian network structure of driver fatigue driving inference is designed,and a digital analysis and judgment model is constructed.By calculating the Bayesian network parameters of voice features,eye features and yawn features,the dynamic Bayesian network is combined After multi information fusion,the specific probability of driver's driving fatigue is obtained,and the summary results are obtained after comparing and analyzing the probability values.Finally,an on-line fatigue driving detection system is designed.The real-time performance is verified based on the platform of Python 3.7.The judgment results of single Fatigue Feature and multiple fatigue features are analyzed.The accuracy of the model is more than 96% verified by the classification confusion matrix.The experimental results show that the system can dynamically adjust the fatigue judgment standard according to the prior probability of driving fatigue,and has good sensitivity and robustness on the whole.