Driver Fatigue Detection Systems: A Review

Tags

Top Highlights

• Chai et al. [62] investigate the effectiveness of Sparse-DBN for EEG based fatigue detection. Sparse-DBN is a semi-supervised learning method.

• Active fatigue is mental depletion caused by active engagement in a task. Humans with intense long hours of work, experience, active fatigue. Passive fatigue is caused by a monotonous task or inattention. Even if a human is not tired a monotonous task will distract from the primary task.

• Fatigue can be classified into active, passive and sleep related fatigue

• A study in [22] has suggested that the sensitivity of driving performance to fatigue is higher on straight roads such as highways as compared to curved roads because driving on highways is monotonous and causes passive fatigue.

• consequently accidents might occur not due to the fact that driver is tired but the driver being distracted from the road.

• The circadian rhythm is a 24-hour sleep/wake cycle and a human will feel sleepy during the same period in the circadian cycle. For adults the largest dip in energy is at midnight (02:00 to 04:00 hours) and midday (13:00 to 15:00 hours)

• Their study suggested that the driver’s manoeuvring ability decreases with increasing fatigue levels and break distance is also dependent on the time of day. As sleep-related fatigue is higher at night (20:00 to 05:00 hours) and early morning (06:00 to 10:00 hours), driving efficiency is compromised during these timings. Philip et al. [25] elaborate on the dependence of driving performance on time of day and time slept in the last 24 hours. If sleep is restricted to two hours in a 24-hour time frame, the inappropriate line crossing increase multi-fold.

• When the duration of driving task increases the steering error and reaction time both increase. After long working hours the driving performance falls significantly

• In 2008 model Crown, Toyota’s first fatigue detection module was installed which detects drowsiness based on eyelid activity. Recently, Toyota has deployed Toyota Safety Sense P [29] in both compact and large vehicles. Toyota Safety Sense P includes vehicle detection, lane deviation, and pedestrian detection.

• Nissan [30] driver attention alert implemented in 2016 Nissan Maxima model tracks the driver’s steering patterns and once it detects any unusual deviation from the pattern a warning signal is generated, in order to alert driver, if break is required. The Nissan driver attention alert adapts to driver’s behaviour by establishing a baseline. Continuously statistical analysis of steering correction errors is performed for the detection of deviation from baseline. A sign is displayed on the console once fatigue is detected.

• Rest Assist in Volkswagen [31] offers lane tracking system, pedal use and erratic steering wheel movements to judge driver fatigue level. Once fatigue is detected the system warns the driver in the form of (i) visual message, (ii) acoustic signal and (iii) steering wheel vibration.

• Smart Eye AB has designed AntiSleep specially for real-time driver fatigue detection. AntiSleep employs features such as eye gaze, head movement, eye position and blinking for driver fatigue detection [32]. Additionally, Smart Eye AB has designed Smart Eye Pro 3.0, which detects fatigue and inattention employing eyelid activity and gaze direction respectively.

• Applied Science Laboratories (ASL) [33] has designed and developed a video-based eye tracking software which observe pupil reflection to measure eye movement.

• Sleep Diagnostics Pvt Ltd designed OPTALERT [34] for fatigue monitoring. OPTALERT utilizes wireless glasses to monitor eyelid and pupil activity. The glasses are equipped with a small-scale LED mounted on frame which measures operator eyelid velocity and eye openness.

• Care Drive’s [35] driver fatigue monitor MR688 incorporates an infrared camera sensor to detect fatigue and tracks pupil changes and head movements.

• The output video is connected to the customers MDVR and fatigue signals are also sent through GPS to the customer so that the supervisor can monitor driver’s state in real time.

• GuardVant has designed OpGuard [36] which is an installable system for driver fatigue detection. OpGuard’s infrared camera monitors driver’s eyelid closure, head and facial movements and behaviour, such as, cell phone use and reading while the vehicle is moving.

• Third party companies have mostly concentrated on fatigue detection using driver physical features such as yawning, blink rate, blink duration and head movement.

• One of the earliest models is the Two Process Model [37]. This model is based on the interaction of two processes namely the circadian Process ‘C’ and the homeostatic Process ‘S’. These processes predict performance and fatigue levels. An upgrade to the two process model is the Three Process Model of Alertness

• he Three Process Model utilizes the duration of sleep and wakefulness as input to predict fatigue risk and alertness. This computer based model considers both circadian and homeostatic component. The Interactive Neurobehavioral Model is targeted for agencies that suggest regulating their working hours, such as aviation, railway and bus transportation. This model is based on homeostatic, sleep and circadian features [39]. The Sleep, Activity, Fatigue, and Task Effectiveness (SAFTE) [40], [41] Model is a neurobehavioral model and works on similar features as the SAFE model. However, the difference between SAFE and SAFTE is that SAFE is based on real life generated data while SAFTE is based on both real life data and data generated by auto sleep algorithm. SAFTE has applications in military, medical and the industrial field. Fatigue Audit InterDyne (FAID) [42] Model only incorporates the hours of sleep as an input and predicts fatigue levels. FAID is applicable to a workplace setup, it associates fatigue levels to past sleep durations.