Prediction of the time when a driver reaches critical drowsiness level based on driver monitoring before and while driving

We collected data on three drivers, on two separate days for each. Once when they slept less than three hours and then with their normal sleep cycle. We recorded their blood pressure, calculation, and reaction times before they started to drive on those days. We then let them drive a cognitive driving simulator while collecting their heart rate variability, driver image information, vehicle information and data for their first minute of driving. We used the data as index of features measured before and during the actual driving, and proposed a method to predict when a driver may reach dangerous state. We evaluated the result of correlation and standard deviation of error between the drivers´ actual critical drowsiness level reaching time and the predicted time obtained by using information before and while driving. From the evaluation, we suggested that we could estimate the time when the drivers´ reach the critical drowsiness level. We also suggested that the heart rate information is not significant enough to increase the accuracy of prediction by observing the changes of the width of the analysis window.