Research On Drowsiness Driving Detection Considering Driving Style

Drowsiness driving is one of the major causes of serious traffic accidents.Research on drowsiness driving detection is of great significance to prevent traffic accidents and improve the traffic environment.However,due to the complex characteristics of drowsiness driving,and the influence of various factors involved in the process of driving,there still exist certain technical bottlenecks in drowsiness driving detection.By the influence of the driver’s characteristics,the existing drowsiness driving detection research found that there are differences in the effect of drowsiness driving detection between different drivers.Relying on Jilin Province talent development fund(Development and Application of Horizontal Support System for Automobile Safety),Jilin Province traffic monitoring and early warning innovation team fund(20130521004JH)and Jilin Province Department of Education "Thirteen Five" science and technology research planning project fund([2016] No.419).Aims at improving the effectiveness of drowsiness driving detection and the suitability of different drivers,this paper analyzes the differences in drowsiness driving characteristics of different drivers,and studies the drowsiness driving detection method considering driving style,from the perspective of the similarity of driver’s driving style behavior.The specific research contents are as follows:(1)Experiment of highway driving style classification.Based on the analysis of the influencing factors of the driving style classification,drivers were recruited and experience was taken in the highway scene,in which drowsiness driving commonly occurred.It was also designed the highway driving behavior questionnaire,and the driving style classification test data were collected.(2)Research on Objective Classification of Driving Style.The highway driving process is divided into three scenarios: free driving,car-following and lane changing.According to the characteristics of different driving scenarios,the corresponding driving style classification index is selected,and the driving style classification index system is constructed.According to the meaning and distribution of each index,the membership function is determined and an objective classification model of driving style based on AHP is established.Drivers were classified based on driving style depend on the experience data,and the driving style classification result is verified by driving behavior questionnaire.The results show that the classification of the driver’s driving style is in good agreement with the driving behavior questionnaire.(3)Experiment of drowsiness driving detection.The driving behavior parameters,operating behavior and vehicle operating parameters of the driver in the normal driving experiment and the drowsiness driving experiment were collected synchronously with the driving simulator and the eye movement instrument.The subjective evaluation method based on face video is used to evaluate the real drowsiness state of the driver,and the data segmentation of the drowsiness driving identification sample is completed,and the database for drowsiness driving detection is constructed.(4)Analysis of Drowsiness Driving Characteristics Considering Driving Style.The waveform characteristics,distribution characteristics and time series complexity characteristics of each parameter were studied by intuitive analysis,statistical analysis and sample entropy comparison respectively.The difference of drowsiness driving characteristics of different driving style drivers were analyzed at the same time.(5)Choice and optimization of drowsiness driving detection indicators considering driving style.Based on the characteristics of drowsiness driving characteristics,the complete set of drowsiness driving characteristic characterization indicators was constructed.According to the validity of single index classification based on the ROC curve for different driving style drivers,the optimal time window of each indicator is determined.Based on the information theory,the relationship between the indicator and the driver’s state,the indicator and the indicator were both considered to select the optimal characteristic indicators subset.(6)The drowsiness driving detection model considering the driving style.Based on the corresponding optimal feature indicator subset and the optimal time window as the model parameters for different driving style drivers,the drowsiness driving detection model considering the driving style is built based on the support vector machine theory.The off-line identification effect of the model was tested by using the drowsiness driving test sample library,a driver drowsiness state monitoring system is designed to verify the online recognition effect of the model.The results show that the drowsiness driving detection model considering the driving style improves the recognition effect on the driver’s state,and solves the problem of general detection model that individual differences in the identification result of drowsiness driving.