related to drowsiness and sleeping behindthe wheel, it is important to acquireprecise knowledge about the extent of theproblem as well as its preconditions andconsequences. The following issues, all of which are investigated empirically in thepresent study, are considered relevant for a better understanding of sleep-relatedaccidents.1. What is the proportion of accidentscaused by fatigue or sleep behind thewheel, and what are the most likely typesof accidents to occur under theseconditions?2. How prevalent is the problem of actuallyfalling asleep while driving?3. What are the most frequentconsequences of falling asleep whiledriving?4. To what extent is sleeping behind thewheel related to characteristics of thedriver, the road and traffic conditions, andthe trip?These drowsiness detection methods canbe categorized into three major approaches:
• Active driving or on
-board detection:Preventing accidents caused bydrowsiness behind the steering wheel ishighly desirable but requires techniques
for continuously estimating driver‘s
abilities of perception, recognition andvehicle control abilities. This paper proposes methods for drowsinessestimation that combine theelectroencephalogram (EEG) log subbandpower spectrum, correlation analysis,principal component analysis, and linear regression models to indirectly estimate
driver‘s drowsiness level in a virtual
-reality-based driving simulator. Resultsshow that it is feasible to quantitatively
monitor driver‘s alertness with concurrent
changes in driving performance in arealistic driving simulator.
• Imaging processing techniques: this
approach analyzes the images capturedby cameras to detect physical changes of drivers, such as eyelid movement, eyegaze, yawn, and head nodding. For example, the PERCLOS systemdeveloped by W. W. Wierwile et. al. usedcamera and imaging processingtechniques to measure the percentage of eyelid closure over the pupil over time.The three-in-one vehicle operator sensor developed by Northrop Grumman Co. alsoused the similar techniques. Although thisvision based method is not intrusive andwill not cause annoyance to drivers, thedrowsiness detection is not so accurate,which is severely affected by theenvironmental backgrounds, drivingconditions, and driver activities (such asturning around, talking, and picking upbeverage). In addition, this approachrequires the camera to focus on a relativesmall area
(around the driver‘s eyes). It
thus requires relative precise camerafocus adjustment for every driver.
• Physiological signal detection
techniques: this approach is to measurethe physiological changes of drivers frombiosignals, such as theelectroencephalogram (EEG),electrooculograph (EOG), andelectrocardiogram (ECG or EKG). Sincethe sleep rhythm is strongly correlatedwith brain and heart activities, thesephysiological biosignals can give accuratedrowsiness/sleepiness detection.However, all the researches up to date inthis approach need electrode contacts on
drivers‘ head, face, or chest. Wiring is
another problem for this approach. Theelectrode contacts and wires will annoythe drivers, and are difficult to beimplemented in real applications.