1.1. Drowsy Driving
Driving is a daily activity for most people in the modern society. Driving while drowsy is dangerous and often leads to accidents [1][2][3]. The National Highway Traffic Safety Administration (NHTSA) estimates that 100,000 police-reported crashes are caused by drowsy drivers each year, resulting in an estimated 1,500 deaths, 71,000 injuries and $12.5 billion in financial losses in the United States. Two of the potential causes of drowsiness-related crashes include: (1) inattention to deviation of the vehicle due to slippery road surface or slight change in steering wheel angle, and (2) not maintaining appropriate distance from other vehicles on the road. In addition, many studies have showed that drowsiness-related crashes often took place during night-time [4], monotonous driving environment [5], or after long hours of driving [6].
Therefore, developing drowsiness detection systems is essential for driving safety.
Several image-based methods have been proposed to monitor the status of the driver or the vehicle. For example, visual cues including eyelid movement, face orientation, and gaze movement (pupil movement) were used to monitor the driver’s vigilance levels [7]. Lane departure warning system (LDWS) or driver assistance system detect lane marking (boundaries) from the video and provide auditory feedback when the vehicle is about to drift off the lane [8][9]. However, monitoring the driver’s status from the video may be affected by changes in ambient illumination and head/body positions or movements. Furthermore, the effectiveness of LDWS is greatly reduced during poor weather conditions or due to unclear lane marking.
Physiological based detection methods, such as electrooculogram (EOG), may be used to detect the driver’s vigilance levels [10]. However, EOG is an indirect measurement of vigilance level and the correlation between EOG and drowsiness is
2
low. Some studies used electromyogram (EMG) to detect the driver’s hand movements [11]. However, EMG is seldom used to detect drowsiness.
1.2. EEG Studies on Drowsiness and Driving
Electroencephalogram (EEG) is one of the most direct and effective physiological measures of arousal states. Changes in EEG power spectra could be used as an indicator for alertness levels. An early EEG study showed increased occipital theta (4-7 Hz) activities as the task performance degraded [12]. Several recent studies have demonstrated the relation between EEG characteristics (e.g., power spectra) and driving performance. Lal and Craig reported an increase in slow wave (theta and delta [0-4 Hz]) activities during fatigue in simulated driving [13]. Schier showed increases in alpha (8-12 Hz) activities during the later laps and replay of simulated driving experiments using the ‘Need For Speed’ PC-game [14]. Campagne et al. showed significant power increases in alpha and theta bands that were highly correlated with the number of running-off-the-road incidents and increase in speed variations [15].
Horne and Baulk showed a correlation between EEG activities in alpha and theta bands and the number of incidents (defined as a car wheel crossing the lateral lane marking) [16]. Lin et al. showed the driving error was positively correlated with EEG log power spectra in the sub-band (< 20 Hz) range [17]. These studies provided the fundamental link between EEG power spectral activities and drowsiness during simulated driving. However, most studies compared the overall EEG power with the mean driving performance in a time window of 30 seconds or longer. In the real life, traffic accidents could occur in a matter of seconds or less if the driver does not promptly respond to sudden events on the road. Therefore, it is essential to investigate the EEG activities before, during and after an event during continuous driving. Huang et al. have demonstrated event-related brain dynamics during
3
continuous performance tasks in a static laboratory setting [18][19][20][21][22].
Decreases or increases in alpha band power occurred following critical events during continuous tracking or driving. However, it is not known whether these event-related dynamic patterns remain the same on a dynamic driving simulator or in real-life driving.
1.3. The Kinesthetic Perception on a Dynamic Driving Simulator
Driving is a complex everyday task that involves predominantly visual information processing. Drivers need to be aware of expected or unexpected critical events (such as deviation of the vehicle) that appear in their useful visual field. In real-life driving, drivers also receive vestibular and proprioceptive inputs, such as vibrations and centrifugal force, in addition to visual inputs. The vestibular apparatus in the inner ear includes the utricle, saccule and three semicircular canals, which provide sensation of balance and head acceleration. Patients with vestibular disorientation syndrome could be a concern for driving safety [23]. Proprioceptor, located in the stretch receptors in the muscles, tendons, and joints, is a sensory receptor that provides information about body position (sense the relative position) and movement of neighboring body parts.Most studies on driving and drowsiness did not consider the influence of kinesthetic stimuli on EEG patterns. In this study, a driving simulator on a six degree-of-freedom motion platform was used to investigate EEG activities from alertness to drowsiness with or without the influence of kinesthetic stimuli when the subjects participated in an event-related lane departure driving task [22].
1.4. Aims of this Thesis
In order to establish the fundamental factors of drowsy driving, the aims of this study were to: (1) draw a comprehensive picture of event-related EEG dynamics (in
4
different brain regions and at different frequency bands) before, during, and after lane departure events, (2) study the influence of kinesthetic stimuli by comparing EEG patterns in motion (active platform) with those in motionless (inactive platform) driving sessions, and (3) identify the brain region(s) and frequency band(s) which could provide useful information to driving safety and drowsiness detection systems.
5