Research On Classification And Detection Of Driving Fatigue Based On Wearable Equipment

In this study,a simulated driving experiment is designed to collect electromyogram(EMG),galvanic skin response(GSR),electrocardiogram(ECG)and eye movement data,judging the driving fatigue level by response time and eye movement indicators objectively.Then,extracting the non-linear characteristic indicators of EMG,GSR and ECG signals,identify different degrees of driving fatigue by support vector machine.Design special traffic scenarios with Silab software,taking the sudden stop of the front car while following it as the stimulus signal,and collect the response time as the main basis for fatigue judgment.Recruit subjects to have simulated driving experiments to obtain EMG,GSR,ECG signal,eye movement data and scene video.By calculating the response time(RT),it is found that it has only two levels,changing to more than 1 second abruptly when fatigue occurs,not increasing gradually,which indicates that fatigue will affect driving behavior when fatigue accumulates to a certain extent.By analyzing blink time,it is found that the distribution of it is multi-longitudinal,which shows that fatigue occurs in intermittent cycles.Combining RT and average every eye-closure time in one minute(CT),taking RT as the first judgment index,followed by CT,propose a driving fatigue classification method,which divides driving fatigue into four grades: awake,mild fatigue,moderate fatigue and severe fatigue.Considering safe driving,moderate fatigue is taken as the appropriate warning stage and mild fatigue is taken as the preventive stage.Extract the characteristic indexes of physiological signals: the complexity of R wave peak(f_C0),the complexity of RR interval time series(RR_C0)of ECG signals,the sample entropy values of EMG(EMG_S)and GSR signals(GSR_S).Analyze the changing trend of characteristic index under different fatigue grades,generally speaking,EMG_S decrease,f_C0,RR_C0,GSR_S increase during fatigue.Significance test shows that there are significant differences in all characteristic indexes under different fatigue grades,and there is no difference in a few pairs.Then,classify and identify four driving fatigue grades by support vector machine(SVM),getting the accuracy rate of 91.41%,except for mild fatigue,the recognition accuracy of each fatigue grade is more than 90%,which shows that the characteristic indexes extracted in this study can effectively identify different fatigue degrees.To sum up,this study innovatively designed the stimulus signal,proposed a method of dividing driving fatigue into four grades according to the demand of safe driving,and determined the optimal early warning time,finally,realized driving fatigue classification through SVM with extracted effective physiological characteristics.It provides a new perspective for driving fatigue classification and lays a theoretical foundation for the development of wearable devices for real-time detection and early warning of driving fatigue.