Article Preview
Top1.Introduction
Drowsy driving is fatal and causes irreparable losses. The U.S.National Highway Traffic Administration has found that 100,000 crashes that happen every year are because of drowsiness at the wheel. Drowsiness has contributed to 15550 deaths, 71,000 fatal injuries, and $12.5 billion monetary losses (World Health Organization, 2009).The AAA Foundation for Traffic Safety estimates that 21 percent of the crashes that occur every year are due to the ‘driver fatigue’ (National Highway Traffic Safety Administration, n.d.). The national sleep foundation has reported that every year 60 percent of adult drivers have been under drowsiness while driving, and 37 percent of drivers had fallen asleep (Asleep at the Wheel, 2010). The main culprits causing drowsiness are prolonged periods of driving, driving alone, night time driving, driver’s anxiety or stress. However, most disappointing factor is that this cause was never pinpointed as a reason for the accidents, since many of the case reports highlight rash driving, alcoholism, and manmade mistakes. With the advent of technology, it is possible to detect drowsiness well before, an accident and prevent the lives. Research on these lines called for a greater focus on designing and development of non-intrusive solutions for detection of drowsiness.
Detection of drowsiness involves identification of certain parameters from the subject or the vehicle. Based on these, the existing methodologies are categorised into Subjective measures, Behavioural measures, Vehicle-based parameters, and Physiological parameters.
Subjective measures of the drowsiness are the driver’s general assessment of the mental status based on a scale Karolinska Sleepiness Scale (KSS) (Dong et al., 2011; Sahayadhas et al., 2012). This method uses general feedback from the driver to give the ratings from 1-9, in which the ratings of 5-9 indicate high Blink duration, large steering wheel corrections, yawning (Ingre et al., 2006). Subjective measures are correlated with the physiological data collected such as EOG and EEG (Hu & Zheng, 2009; Portouli et al., 2007). Subjective assessment is not possible to implement in real-time scenario.
Behavioural measures of the driver include the change in the position of the head, pupil diameter, blink duration, yawning, gaze, which are monitored by several sensors (Dong et al., 2011; Sahayadhas et al., 2012). One method uses Eye tracker employing one or multiple cameras to monitor the pupil and the line of vision of the driver in the vehicle. This device continuously tracks the eyelids, and if the eyelids are closed for more than a pre-defined limit, an alarm rings to alert the driver (Husar, n.d.). A Non-intrusive real-time image acquisition system with efficient image processing algorithm was developed to detect the vigilance level by Eye, Gaze, Face and Pose tracking (Ji & Yang, 2002). These methods (Bergasa et al., 2006; Husar, n.d.; Ji & Yang, 2002) have a limitation as a camera cannot be used under the poor illumination conditions. PERCLOS the Percentage closure of the pupil over time is measured in many studies (Abe et al., 2011; Lexus LX Driver Monitoring System, n.d.; McKinley et al., 2011; Seeingmachines Driver State Sensor, n.d.) and is commercially available as Seeing Machine and Nexus. These technologies employ High Definition IR camera to capture the eye and apply efficient image vision technologies to find the PERCLOS. Several other methods have used yawning and position of the head as the criteria (Murphy-Chutorian & Trivedi, 2010; Nature, n.d.; Smith et al., 2003; Xue et al., 2009).