Driver Discrepancy Elimination Model And Its Application In Driver’s Mental Load Evaluation

With the gradual improvement of people’s quality of life and the rapid development of the automotive industry,the number of automobile is increasing,and there are more and more traffic accidents.In all traffic accidents,the driver’s own factors account for more than ninety percent of all numbers.During the driving process,the driver needs to comprehensively analyze various aspects of the car’s driving state,road conditions,etc.,so it bears the corresponding mental load.When the mental load of the driver is too high or too low,it is easy to cause decision-making mistakes,leading to accidents.Under such a huge car scale,the differences between drivers become more and more obvious.Due to the differences between individuals,different drivers bear different mental loads when facing the same situation.Most of the current driver status studies only consider the impact of subtask information on the driver,and do not consider the impact of differences between different drivers on the driving assistance system.Therefore,this paper focuses on the impact of subtasks on the driver status,and uses feature engineering technology to remove the differences between drivers and select features with high correlation to identify the driver’s mental load status under the subtask.The main research work of this paper is as follows:First,a detail summary of the researches on mental load of drivers was elaborated,analyzes the current commonly used evaluation methods,and proposes the importance of removing differences between drivers.The driver’s mental load induction experiment is designed,and by setting up a driver simulation platform and a virtual experiment scene,the driver’s EEG(Electroencephalogram)data of different load levels under different tasks are collected.Secondly,a Driver Discrepancy Elimination model is proposed.To address the noise problems in EEG signals,the model uses independent component analysis first to remove the EEG noise in the EEG signals.Then,extract the features of the time domain,frequency domain,and time-frequency domain from the EEG signals,and use feature projection technology-nuisance attribute projection method removes the differences between drivers in the EEG signals.Finally,the feature selection methodFisher discriminant ratio is used to select features with higher correlation in the EEG signals.After the model,the EEG data features can better reflect the driver’s mental load state,and between different states,can better remove the impact caused by the differences between the drivers.Finally,the BP neural network classifier are debugged to identify driver’s mental workload.It uses the feature projection technology to remove the driver difference and the feature selection method with high correlation features.The effectiveness of the Driver Discrepancy Elimination model was verified by setting four sets of comparison experiments.Group experiments selected the best number of features for feature selection.