| Popis: |
Accident statistics show that driver drowsiness has been one of the major cases of road accidents. Therefore, a system detecting driver’s drowsiness and recommending a driver to take a rest is expected to be effective to reduce number of accidents and fatalities. There are different methods proposed for such a system. In general, using lane information detected by front observing camera is expected to reach high accuracy of drowsiness detection, but the drawback relying on the detected lane information is weakness of the system due to no lane makers or weather condition. We propose in this paper a method based on vehicle’s inertial sensors and driver’s steering operation, but not relying on lane information by camera. In contrast to other systems in market, we have found that steering vibration additional to acoustic warning is effective to warn a driver in drowsy condition. One of novelty of the proposed method is that it has taken the individual difference into account by identifying parameters of a driver model in the algorithm. Other influence due to driving situation causing wrong drowsiness estimation has been analyzed and countermeasures for these effects are also integrated in the proposed method. Since the proposed method is able to estimate five levels of driver’s attention, a driver is able to see the attention level decrease with the attention level meter in display while driving. We confirmed in experiments that the proposed method has sufficient accuracy and low false positive rate. Especially the experiment shows that false positive rate is 0.14 times per hour. |
