Study On Driving Distraction Based On Rhythm Physiological Signals

Traffic safety has always been one of the most serious problems in modern society.Statistics show that with the increase in car ownership,our country has become one of the countries with the highest number of fatal traffic accidents in the world.In our country,the annual economic losses caused by traffic accidents are more than 1 billion yuan,of which driver-related traffic accidents account for about 90% of the total accident rate.As the most important part of the closed-loop system of people vehicles and road,human factors are more and more concerned by relevant scholars.How to get the rhythmic physiological data under the driver distraction state from this point of view so as to propose an effective prediction model to reduce the incidence of traffic accidents and the casualty rate caused by driver factors is the focus of this article.This paper first introduces the academic definition of driver-induced traffic accidents classification,driving disorders include: fatigue driving,drunk dri ving,emotional disorders and distraction,the traditional method of recognition of driving disorders is based on vehicle data or video data,by Limited to the system's robustness and economy have not been well applied.According to the observation of the market,we found that the acquisition of human physiological data is becoming more and more reliable with the vigorous development of smart wearable devices.The correlation between human physiological data and driving disorders has been proved through literature research.If we can establish a prediction model and gives a reliable prediction of driving obstacles so as to make corresponding emergency treatment before the accident,so that the above objectives of enhancing traffic safety can be achieved to some extent.Predicting driving distraction using the driver's rhythmic physiological indices is a problem of pattern classification and can be solved by machine learning algorithm modeling.In order to acquire the physiological data of driver's rhythms use d to establish the driver's distraction recognition model,a specific driving simulation experiment needs to be performed.In the stage of experimental design,a virtual environment of driver simulation experiment was set up by independently.Then orthogonal experimental design was used to explore the external conditions that had a significant impact on driver distraction.The optimal experiment Combination of conditions.Based on the combination of the optimal experimental conditions,the five rhythmic physiological data of 20 volunteers under driving simulations were obtained by using the comparatively mature dual task paradigm of human factors engineering.Based on the driver's physiological data obtained from the above experiments,machine learning modeling is proposed.In the process of machine learning modeling,firstly,the EEG(EEG signal)is used as the training sample to model and compare the traditional training machine learning methods SVM(Support Vector Machine)and CNN(Convolutional Neural Network)After discovering the advantages of CNN in dealing with this sort of problems,we continue to carry out the end-to-end training of other signals based on the Le Net-5-based convolutional neural network model and evaluate its performance.Finally,a framework of real-time driver intelligent detection based on the driver distraction prediction model is proposed.