Research On The Evaluation Parameters And Model Of Driver’s Attention During Distracted Driving

With the continuous growth of China’s automotive vehicle ownership and the rapid development of road traffic industry,frequent traffic accidents cause property and life safety problems to drivers.It is reported that 30% of traffic accidents are caused by distracted drivers.In the above traffic accidents,the proportion of traffic accidents caused by the driver’s active participation in secondary tasks is as high as 36.4%,and distracted driving has become one of the important causes of traffic accidents.The external characteristics of distracted drivers are not as uniform as fatigue driving.On the contrary,drivers’ distraction duration is short,greatly influenced by driving environment,and their external characteristics are diverse.Therefore,it is of great significance for the development of road safety driving state detection.In this paper,distracted driving is taken as the research object,and the external characteristics and detection algorithm design of driver’s visual distraction and cognitive distraction in highway scene are carried out through driving simulator.This paper systematically reviews the research status of distracted driving detection at home and abroad.The detection method,evaluation index and existing products of distracted driving are introduced in detail.Based on the current lack of research,aiming at the current situation of poor systematicness between feature indexes and imperfect research on visual parameters,a framework of distracted driving detection algorithm is proposed.The experimental process of distracted driving is designed.After analyzing the closed-loop system of human vehicle and road,this paper proposes to evaluate the driver’s attention state from two dimensions of vehicle dynamic parameters and driver’s eye movement parameters.After sorting out the factors that can affect the distracted driving,the experimental object,equipment parameters,road environment design and experimental process are designed and introduced in detail.The problem of data stream synchronization is solved by means of integrated development.For the collected data,through Matlab＿2015a、Origin＿2017 and other data processing software,quartile,MaxMin normalization and other methods are used to process the data,so as to prepare for the data analysis of the following papers.Analysis of vehicle operation parameters.This paper analyzes the influence of distracted driving on average speed,speed standard deviation,acceleration standard deviation,steering wheel angle and steering wheel angle entropy from two dimensions of vehicle longitudinal parameters and lateral parameters.Based on the analysis of variance,3 key parameters,speed standard deviation,steering wheel angle and steering wheel angle entropy,which have significant influence on driver’s attention state,are obtained and used as parameters to identify the driver’s attention state.Analysis of driver’s eye movement parameters.Firstly,the main eye movement forms of drivers are analyzed,and the distinguishing standard of gaze and saccade is established.Through the analysis of variance,4 kinds of parameters which have significant influence on driver’s eye movement parameters are obtained from fixation,saccade and blink.Finally,the entropy rate of fixation area,standard deviation of horizontal sight,standard deviation of vertical sight and average saccade speed are determined as the parameters to estimate driver’s attention state.The driver attention state detection algorithm is designed.Pearson correlation coefficient is used to test the correlation between the parameters.Then,based on the feature that LSTM can process the time series information of multivariable,the detection model of this paper is constructed,and the key parameters are introduced in detail.Then,the driver’s attention state detection algorithm is proposed,and the sliding window decision module is designed according to the decision sequence results.In order to deal with the two kinds of possible errors,a discriminant verification module is proposed.The confusion matrix is used to calculate the accuracy,precision and F score of the algorithm to evaluate the performance of the model.Based on the driving simulator experiment,the classification results of the algorithm are verified,and the influence factors of the algorithm are discussed.The results show that visual distraction has a more significant effect on drivers than cognitive distraction.Considering the accuracy of the algorithm and the rapidity of detection,the best combination of algorithms is when the length of time window is 5s and the coincidence of time window is 75%.In addition,under the same driving conditions,experienced drivers can take more compensation measures to ensure safe driving when dealing with distracted driving.