Research On The Recognition Method Of Drinking Driving State Based On Individual Driving Behavior

In the traffic system, the most important component element is driver, which hasthe leading role for vehicle travelling. Once there are problems with drivers and theirdiriving ability decline, their driving safety will be threated. The statistics of roadaccidents also showed that most of the accidents were caused by drivers themselves.The inappropriate driving behaviors of them are the important reasons for theaccidents. Therefore, the study on drivers is very important on the researches of trafficsafety. Especially, the study on the correlation between driving behavior and trafficsafety has great significance.Drinking driving is one typically dangerous driving state. It has high probabilityto lead to terribleaccidents. Although drinking driving has been prohibited withenacting laws, there are still drinking drivers. Therefore, new recognition method ofdrinking driving state needs to be studied to assist the detection of drinking drivers.Driving behavior, which is continuous parameters of drivers in vehicle controlling, isdirect affected by drivers’ driving states. It also can be detected without contact withdrivers.The recognition based on driving behavior provides a new method for thedetection of drinking driving. However, individual difference is an important factor todecline the accuracy of driving behavior study. It needs to be resolved firstly and thenthe recognition method based on individual driving behavior is presented to improvethe recognition accuracy.In this study, the relation between driving behavior and drinking driving statewas researched based on driving simulation synthetic experimental platform. Theeffect mechanism of alcohol on drivers was analyzed. The basic characteristics ofdriving behavior under the influence of alcohol were studied. The model of individualdriving behavior was built for each driver. Followed this model, the recognitionmethod of drinking driving state based on individual characteristics was presented.The main contents studied are as follows:Firstly, the comprehensive experiment was designed based on the drivingsimulation synthetic experimental platform to collect data. The synthetic experimentalplatformhuman factor oriented was introduced according to the study and method fordrivers. Encompassing the driving simulator,the platform was built integrating withsubjective questionnaire, driving adaptive characterisitics detectors and dynamic physiological detectors. The platform can used to collect drivers’personalitycharacteristics, driving adaptive characterisitics and driving characteristics. Based onthis platform, the experiment was designed to acquire experimental data of drivers innormal driving states and drinking driving states with different drinking levels. Thescenarios and experimental procedure were introduced. The controlling method foreach factor in the experiment was elaborated. The basic data were collected from theexperiment to support this study.Secondly, the effect mechanism of alcohol on drivers was analyzed. The analysisfocused on the subjective characteristics and driving adaptive characteristics ofdrivers in their drinking driving states. The subjective characteristics’ distributionswere analyzed with statistical method to show the subjective change features. Thedriving adaptive parameters were studied with repeated ANOVA and logisticregression method to explain the effect mechanism of alcohol on drivers. The resultsshowed that when drinking driving, drivers are impulsive and have sensation seekingattitude, and their cognitive reaction ability decline. What is more important,drivers’perception ability is impaired significantly. The ability of acquiring roadinformation correctly declines, especially for visual sense. According to these results,the method to select driving behavior indices was discussed. The indices includedvehicle’s travelling state parameters and driver’s operation behavior parameters.Thirdly, the driving behavior characteristics under the influence of alcohol wereanalyzed. The driving behavior data in normal driving and drinking driving states withdifferent drinking levels were analyzed with statistical method. In the first place, thevehicle’s travelling state parameters, speed and lane position were analyzedconsidering different road conditions. The results showed that when drinking driving,the speed is higher and the stability of speed and lane position declines. It was alsoproved that the effects of alcohol on driving behavior are related with road conditions.In the second place, the drivers’ operation behavior parameters, accelerator pedal,brake pedal, steering wheel and clutch were studied. The results indicated that underthe influence of alcohol, drivers stamp on the pedals deeper and faster, release theclutch faster and react slowly.At the same time, under the influence of alcohol, drivers’small operation of steering wheel reduces and the regularity is lower.According tothese results, some driving behavior indices affected significantly by drinking drivingwere extracted to explore the recognition method of drinking driving state based onthe basic characteristics of driving behavior. Fourthly, the individual characteristics of drviers were verified and the individualmodel of driving behavior was built. Correlation analysis was used to study drivingbehaviors of each driver himself and aomong different drivers to explain individualcharacteristics and individual differences. Based on the time sequence feature ofdrivinig behavior, Hidden Markov Model (HMM) was used to model driving behavior.The results indicated that individual characteristics come from each driver’s drivinghabits and the modelling according to time sequence feature is feasible. Then thedriving behavior of drivers in each typical road was modeled withGaussian mixtureHidden Markov Model (GHMM), and its effectiveness was verified.At last, the recognition method of drinking driving state based on individualcharacteristics was presented. The models’ parameters of individual driving behaviorsfrom normal driving and drinking driving were contrasted to find the change tendency.Then6indices were extracted. Based on these indices, the recognition method ofdrinking driving state was presented and its recognition potential was verified. At thesame time, the software platform of harmful driving evaluation based on drivingbehavior was introduced. The primary algorithm of the recognition and warningmethod of drinking driving state in the platform was explained.In this study, the analysis of driving behavior characteristics was the key point.Individual driving behavior model was built for each driver to resolve the individualdifferences problem. A new recognition method of drinking drivinig state waspresented according to the effect mechanism and the relation between drivingbehavior and driving state. The results of this study are not only helpful for thedrinking driving problem, but also promote the studies on driving behavior.