Using Spectral Analysis to Extract Frequency Components from Electroencephalography: Application for Fatigue Countermeasure in Train Drivers
Budi Thomas Jap*, Sara Lal*, Peter Fischer+, Evangelos Bekiaris^
University of Technology, Sydney (UTS), Broadway NSW 2007, Australia
+ Signal Network Technology, Lane Cove NSW 2066, Australia
^ Center for Research and Technology Hellas, 6th km Charilaou-Thermi Road, 57001 Thermi, Greece
[email protected], [email protected], [email protected], [email protected]
AbstractTrain accidents can have a massive impact towards the surrounding area as well as the general community. Most train
accidents can be attributed to fatigue, and hence, development of fatigue countermeasure devices that can warn drivers of fatigue status and prevent accidents can greatly benefit train drivers, passengers, society and general community. Electroencephalography (EEG) has been proven to be one of the most reliable indicators of fatigue. This study investigated the change of brain activity during fatigue-instigating monotonous driving session, by extracting the four frequency components (delta, theta, alpha, and beta) using FFT spectral analysis at different brain sites (frontal, central, temporal, parietal, and occipital). Results identified some statistically significant differences between early and later stages of driving in delta, theta and beta activities at different brain sites. The results of the current study may be used for future development of fatigue countermeasure by targeting specific frequency component and brain sites
Keywords: Spectral analysis, fatigue, countermeasure, electroencephalography
1. IntroductionRailway is the backbone of most industries to transport goods from a factory or warehouse to the
consumers in other states, and it is one of the cheapest methods for people to travel interstate and
intrastate [1]. The number of train accidents and injuries in Australia is relatively low, and compared to road
and aviation industries, rail is a relatively safer transportation means [2]. However, given the fact that the
travelling speed of a train is relatively faster than a car, a train carriage is much heavier than a car, and that
trains’ braking time is slower than a car (about 1.1 m/s2) [3], the impact that it creates in an accident can
greatly affect the surrounding area and the community as a whole. Bureau of Transport and Regional
Economics (BTRE) [4] reported that not only the people directly involved in the accidents, such as the
drivers and the passengers, but also the whole community is affected by the accident.
Fatigue is found to be one of the three biggest killers on Australian roads, along with speeding and drink
driving, and research indicates that fatigue is four times more likely the cause of an accident than drugs or
alcohol [5]. National Transport Commission (NTC) [2] has acknowledged fatigue as a significant causal
factor in industrial accidents, especially in those that operate on a continuous basis and require workers to
work long hours, night shifts or rotating shifts. Transportation industries, such as rail, road, and so on, all
face similar challenges to minimize the degree of fatigue in its workforce [2].
Approximately 75% of all train accidents can be attributed to human errors, which are mostly caused by
train crew members when they are in a fatigued state [6, 7]. Train driving task is a complex task and relies
heavily on several aspects of neuron-cognitive functioning, including sustained attention, object detection
and recognition, memory, planning, decision-making and workload management [8]. Fatigue affects both
the physical capacity and the cognitive and other mental processes to perform work [2], and is associated
with lapses in attention, longer response times and more frequent errors, and also have increased difficulty
identifying and processing important information from the surrounding environment [8]. For example, it has
been noted that train drivers in a fatigued state tend to use the brake less and travel at faster speeds [8].
Therefore, there is a need for a robust fatigue countermeasure device to detect when the driver starts
showing early signs of fatigue. There are many developments of fatigue countermeasures, such as
development of electroencephalography (EEG) algorithms to detect fatigue [9], facial movement detector
[10], and PERCLOS, which detects the percentage of eye closure [11]. However, Artaud et al. [12] found
that EEG is one of the most reliable indicators of fatigue. A study by Lal and Craig [13] found a good test
and retest reliability and high reproducibility for the delta and theta bands, and another study by Gasser,
Bacher and Steinberg [14] found the frequency of alpha rhythm also showed a good reliability. Acceptable
test and retest reliability has also been found in resting alpha activity [15], and others found acceptable
reliability for theta, alpha, and beta frequency bands [16].
Four frequency components in EEG exist, which are delta (δ) (0 – 4 Hz), theta (θ) (4 – 8 Hz), alpha (α)
(8 – 13 Hz), and beta (β) (13 – 35 Hz), which can be measured to detect the current state of a driver [17].
Delta activity shows higher activity during sleep. An increase in theta activity can indicate early stage of
drowsiness [18]. Alpha activity indicates a state of relaxed wakefulness, which decreases with
concentration, stimulation, or visual fixation [19]. An increase in alpha activity has also been found in train
drivers, who were really sleepy and fell asleep while driving [18, 20]. An increase in beta activity is related
to alertness level, and a decrease with drowsiness [21]. Torsvall and Åkerstedt [20] believed that alpha
activity was the most sensitive measure that could be used in detecting fatigue, followed by theta and delta
activities. However, delta activity is more related to occurrence of sleep proper [20].
Several studies have proposed methods of fatigue detection using EEG, such as a study by Tietze [22]
that proposed the detection of alpha spindles to detect fatigue. Other studies have proposed two
algorithms, which were (θ+α)/β and β/α, that can be used as a fatigue detection technique [21, 23]. Others
have proposed the combination of FFT and neural network to classify alertness and drowsiness [24], the
use of wavelet transform [25, 26], or the use of Independent Component Analysis (ICA) algorithm [27].
The current study investigated the changes of EEG activities, delta, theta, alpha, and beta, during
fatigue-inducing monotonous driving session for future development of fatigue countermeasure devices.
2. Methods
A total of 52 non-professional drivers (36 males and 16 females), aged 20 to 70 years (mean 26 9
years) were recruited to perform a monotonous driving simulator task. All participants provided informed
consent prior to participating in the study. For selection criteria, The lifestyle appraisal questionnaire was
used as a guideline that required all participants to have no medical contraindications, such as severe
concomitant disease, alcoholism, drug abuse, and psychological or intellectual problems that were likely to
limit compliance [28].
The study has approval from the Human Research Ethics Committee, and was conducted in a
temperature-controlled laboratory at around noon 1.5 hours. Participants were asked to refrain from
caffeine consumption (tea or coffee) and from smoking approximately 4 hours prior to the study. Alcohol
consumption was also refrained for approximately 24 hours before the study. All participants reported
compliance with these instructions.
Grand Prix 2 (version 1.0b, 1996, Microprose Software, Inc., USA) was used as the driving simulator
software for the study. The display from the software showed other cars, the driving environment, the
current speed, and other road stimuli. The simulator consisted of a car frame with an in-built steering
wheel, brakes, accelerator, and gear change buttons.
Two driving sessions were completed for the purpose of the current study. The first driving session was
the alert or active driving session for approximately 10 to 15 minutes, which would serve as a baseline
measure. During this driving session, the driving environment involved other cars and stimuli on the road.
The second driving session was the monotonous driving session for approximately 1 hour with speed limit
restricted to between 60 – 80 km/h. This session involved the participants driving with very few road stimuli.
Simultaneous physiological measurements were recorded during the driving sessions, using NeuroScan
system (Compumedics, Australia). A total of 30 channels of electroencephalography (EEG) were
measured simultaneously while participants were driving. The 10-20 international standard of electrode
placement was applied [29]. A referential montage was used with the reference point located at the center
of the head, between the midline central electrode (Cz), and the midline central parietal electrode (CPz).
The sample rate for the EEG recording was 1000 Hz. The current NeuroScan system was not a wireless
EEG system, and was only used for data collection and analysis purposes, prior to the development of
simple and wireless EEG system that was viable for use in the real driving environment. In addition to the
EEG recording, blood pressure and heart rate were collected before and after the driving task.
The literature specified that delta frequency range was from 0 – 4 Hz [17], and it is easily affected by DC
voltage (0 Hz), which might compromise the result of the experiment. Therefore, the EEG time-domain
recording was filtered with a 6th order Butterworth high-pass filter with cutoff frequency at 0.5 Hz.
Butterworth low-pass filter was also applied with cutoff frequency at 35 Hz to filter out frequencies higher
than beta frequency band, such as muscle artifacts.
All 30 channels of the EEG recording were then split into 1-second epochs, and were subjected to Fast
Fourier Transform (FFT) algorithm (Hanning Window) to derive spectral power for the four frequency
components, which were delta, theta, alpha, and beta. The total data for the monotonous driving session
was segmented into 10 equal time sections of approximately 6 minutes for each section. Three lots of thirty
1-second epochs were averaged in each of the alert and the monotonous driving sections to obtain on
value for each section. The result for the ten sections of monotonous driving session was then compared to
the alert baseline. The EEG channel values for each section (both in alert and monotonous driving
sessions) were then averaged according to the different sites of the brain, where the electrodes were
located, to obtain 5 site averages, which were frontal, central, temporal, parietal, and occipital sites.
For the current study, delta, theta, alpha, and beta activities were compared during the data analysis to
identify changes in those activities during fatigue-inducing driving session. Analysis of Variance (ANOVA)
was performed to identify significant differences between the 10 time points and the alert baseline. This
analysis was performed for delta, theta, alpha, and beta separately for the five different sites of the brain
(central, frontal, occipital, parietal, and temporal).
3. Results
The average body mass index (BMI) of participants was 23 ± 7 kg/m2 (normal range: 18.50 – 24.99
kg/m2 [31]), which indicated that the participants were in normal proportion between weight and height. The
average pre-study blood pressure was 118 11 mmHg (systolic) and 75 9 mmHg (diastolic), and the
average post-study blood pressure was 114 13 mmHg (systolic) and 71 17 mmHg (diastolic). The
average heart beat was 72 10 beats/sec for the pre-study and 65 9 beats/sec for the post study. T-
test analysis on the pre-study and post-study results of blood pressure and heart beat revealed significant
differences for systolic blood pressure and heart beat (refer to ). Heart beat and blood pressure are
normally higher when a person is alert and lower during fatigue or resting period [32]. The statistically
significant lower blood pressure and heart beat at the end of the study showed that participants were
considerably fatigued after the monotonous driving session. The average driving time was 1 hour and 3
minutes 12 minutes. Gillberg, Kecklund and Åkerstedt [33] has previously demonstrated that a 30-
minute of monotonous driving will lead to lower alertness level, which is reflective of a fatigued state.
Table 1 t-test result for difference between pre- and post- measurements of blood
pressure and heart beat (significant p-value italicised)
Diff Std. Dev. Diff t p
Systolic BP 4.1 10.92 2.8 0.009
Diastolic BP 4.6 16.83 2.0 0.05
Heart beat 5.9 8.025 5.4 < 0.001
Note: BP = blood pressure
Figure 1 shows an example of brain topography for one of the volunteers, indicating delta, theta, alpha,
and beta activities. Red-shaded areas indicate high activity, and blue-shaded areas indicate low activity. In
this example, delta activity increased in the frontal region at the beginning of the driving session (sections 1
to 5), and decreased towards the end of the driving session (sections 8 to 10). Unlike delta activity, theta
activity steadily increased in the frontal region, starting from time section 3. Theta was higher in the frontal,
temporal, and occipital areas towards the end of the driving session (sections 9 to 10). Alpha activity in the
occipital region increased gradually, starting from low alpha activity in section2, to high alpha activity in
section 10. Beta activity showed almost the opposite from the alpha and theta activities, in which it showed
a decrease of activities. Beta activity showed higher activity in the frontal and temporal regions during alert
driving, and decreased significantly towards the end of the driving session (starting from section 6).
Figure 1 Example of brain topography of alert baseline and 10 time sections during
monotonous driving
ANOVA analysis revealed significant differences during the comparison between the alert baseline and
the 10 time points of monotonous driving. Significant differences were found at temporal sites only for alpha
(F = 2.0, p = 0.03), and beta (F = 2.3, p = 0.01). No significant differences were found at other sites. Delta
showed significant differences for the frontal site (F = 3.5, p < 0.001), whereas theta showed significant
differences for central (F = 2.3, p = 0.01), frontal (F = 2.3, p = 0.01), and parietal (F = 3.6, p < 0.001). Refer
to Table 2.
Table 2 ANOVA result for delta, theta, alpha and beta
C F O P T
Delta - 3.5/<0.001 - - -
Theta 2.3/0.01 2.3/0.01 - 3.6/<0.001 -
Alpha - - - - 2.0/0.03
Beta - - - - 2.3/0.01
Note: Showing significant p-value only (<0.05). C = central, F = frontal, O = occipital, P = parietal, T =
temporal.
Beta activity revealed significant difference between alert baseline and time section 10 (p = 0.031). Post-
hoc Bonferroni analysis for alpha activity revealed no significant differences. Post-hoc Bonferroni analysis
for delta activity at frontal site revealed significant differences in activities between alert baseline and time
sections 1 (p < 0.001), and 2 (p = 0.024), and between time sections 1-4 (p = 0.017), 1-5 (p = 0.021), 1-8 (p
= 0.008), 1-9 (p = 0.036), and 1-10 (p = 0.032). Post-hoc Bonferroni analysis for theta activity for central (p
= 0.006), and frontal (p = 0.006) sites revealed significant differences between alert baseline and time
section 2, whereas result for parietal site revealed significant differences between alert baseline and time
section 2 (p < 0.001), and between time sections 2-4 (p = 0.028), 2-5 (p = 0.015), 2-6 (p = 0.037), 2-7 (p =
0.048), 2-8 (p = 0.035), 2-9 (p = 0.023), and 2-10 (p = 0.021). Refer to Table 3.
Table 3 Post-hoc Bonferroni analysis for results showing significant p-value in ANOVA
Site Comparat
or
01 02 03 04 05 06 07 08 09 10
Delta F Alert <0.001 0.02 - - - - - - - -
Delta F 01 - - 0.02 0.02 - - 0.008 0.04 0.03
Theta C Alert - 0.006 - - - - - - - -
Theta F Alert - 0.006 - - - - - - - -
Theta P Alert - <0.001 - - - - - - - -
Theta P 02 - - 0.03 0.02 0.04 0.048 0.04 0.02 0.02
Alpha T - - - - - - - - - -
Beta T Alert - - - - - - - - - 0.03
Note: Showing significant p-value only (<0.05). C = central, F = frontal, P = parietal, T = temporal.
The plots shown in Figure 2 show an example of the changes for delta, theta, alpha and beta activities
over time at the temporal site. The slow wave EEG activities (delta and theta) show an initial increase
followed by a decrease, and then is steady over time (Figure 2a and b), whereas the fast wave EEG
activities (alpha and beta) show a general decrease over time, indicated by a slight decrease of alpha
activity, and a steeper decrease for beta activity (Figure 2c and d).
Figure 2 Temporal activity plotted over time during driving for delta, theta, alpha, and
beta
4. Discussion
There is strong evidence supporting a relationship between railway accidents and locomotive crew
fatigue, such as time on duty, work-sleep-rest cycles and shift work [6]. Fatigue and severe sleepiness at
work are common among personnel responsible for train driving and the control of train traffic [34, 35].
Härmä et al. [34] conducted a study on irregular shift and the effect on sleepiness for train drivers and
controllers and found that over half of the train drivers and traffic controllers reported severe fatigue during
their night shifts, and fatigue occurred in about 20% of all morning shifts. Austin and Drummond [35] found
that about 25% of train drivers dozed off while driving or waiting at the station. Unlike the driver of a car
who can swerve his or her vehicle to avoid collision, a train driver can only apply the emergency brake,
sound the horn and hope that the train will stop in time before the collision [35]. Hence, investigation into
possible means to prevent train accidents will greatly assist the community.
EEG has been found to be a reliable fatigue indicator [12] and can be used in a fatigue countermeasure
device. It also has been proven to have a high reliability and reproducibility in the delta, theta and alpha
bands in the previous studies [13, 14]. Hence, this study assesses the change in brain activity from the
alert baseline to the monotonous driving session.
From the result of the current study, significant differences were found at the temporal site for alpha and
beta activities, at the frontal site for delta activity, and at the central, frontal, and parietal sites for theta
activities. The EEG activity in alert baseline and early stages of fatigue were significantly different to the
later stages of fatigue. For example, the delta activity in the frontal site, theta in the parietal site, and beta
activity in the temporal site, which were significantly lower during later stages of driving.
Figure 2 captures the information presented in Table 2 and Table 3 for the temporal site only. Other sites
of the brain also show similar changes for each brain activity, with slight difference in the magnitude of
change. However, from this figure, a significant decrease in beta activity can be observed, which can also
be utilized in a fatigue detection algorithm. The gradual decrease of alpha activity can also be utilized
together with beta activity to improve robustness of fatigue detection process.
Different EEG algorithms can be combined together in order to increase reliability and accuracy when
detecting fatigue [36]. This approach would most likely be successful for on-road fatigue detection in the
longer term, because it combined different approaches to produce a solid fatigue detection system.
However, such system can become complex because there will be some needs for an agreement in the
outcome of the analysis of all the different algorithms about the current state of the driver. The
disadvantage of using multiple algorithms in detecting driver fatigue is when the outputs contradict each
other and the system does not have certainty whether a driver is fatigued or not.
However, as reliable as it can be in detecting fatigue, the setup for fatigue monitoring process using an
EEG is a complex procedure, and in its current status, it would be difficult to use in the real driving
environment [37]. Hence, simple and wireless EEG systems that can be worn by drivers whilst driving need
to be researched and developed, before fatigue countermeasures employing the accuracy and reliability of
EEG to detect fatigue will become viable.
For the purposes of transmitting an alert signal to the train dashboard, Bluetooth can be utilized as a
Personal Area Network (PAN) for the application of a simple EEG device to monitor fatigue. Bluetooth is a
non-expensive, short range data communication with low data rate and low power consumption [38-41].
The fatigue countermeasure device needs to be battery-powered, and hence, Bluetooth’s low power
feature works together to prolong the battery life of the device. Train drivers will be confined to the train
cabin, and hence, the short range data communication that Bluetooth offers is sufficient for the fatigue
detection application. The short range data communication can also be viewed as a security measure,
which reduces the probability of eavesdropping from other wireless devices.
The existing communication network in the current train system can also be utilized to transmit
information back to the train control base. Should the train be stopped due to the driver dozing off, or
having other problems, during driving, the control base may need to be alerted to this event, which may
enable them to take appropriate safety measure to the situation.
5. Conclusion
The result of the current study showed a significant change in the brain activity during a fatigue-inducing
driving session, and several sites that showed significant changes during fatigue could potentially be used
to detect fatigue. Future development of EEG-based fatigue countermeasure device can utilize the results
of the current study to improve detection of fatigue by targeting specific frequency bands and sites that
show significant changes. However, because of the complexity of the current equipment, the equipment
used in this study was not suitable for use in future driver fatigue countermeasure device. A simple and
wireless EEG device needs to be developed to suit the driving environment. The EEG device may utilize
sites on the brain, at which significant differences occur between alert and fatigue states in this study, such
as a large drop of beta activity over a period of time. This study has implications on future development of
driver fatigue countermeasures using wireless technology for application in the railway industry.
6. Acknowledgement
The research was supported by an ARC Linkage grant Australia (LP0560886) and by SENSATION
Integrated Project (FP6-507231) co-funded by the Sixth Framework Programme of the European
Commission under the Information Society Technologies priority.
7. References
[1] RailCorp, "Rail Corporation New South Wales, 2004-05 Annual Report," NSW, Australia 2005.
[2] National Transport Commission (NTC), "Fatigue Management Within The Rail Industry: Review of
Regulatory Approach," Melbourne, Australia 2004.
[3] M. Ashiya, S. Sone, Y. Sato, and A. Kaga, "Application of pure electric braking system to electric
railcars," in Proceedings of 6th International Workshop on Advanced Motion Control, 2000, pp.
163-168.
[4] Bureau of Transport and Regional Economics (BTRE), "Rail accident costs in Australia,"
Department of Transport and Regional Services, Commonwealth of Australia, Canberra 108, 2002.
[5] Parliamentary Inquiry, "Beyond the midnight oil: An inquiry into managing fatigue in transport,"
House of Representatives Standing Committee on Communication, Transport and the Arts,
Canberra 2000.
[6] G. J. S. Wilde and J. F. Stinson, "The monitoring of vigilance in locomotive engineers," Accident
Analysis & Prevention, vol. 15, pp. 87-93, 1983.
[7] G. D. Edkins and C. M. Pollock, "The influence of sustained attention on Railway accidents,"
Accident Analysis & Prevention, vol. 29, pp. 533-539, 1997.
[8] J. Dorrian, G. D. Roach, A. Fletcher, and D. Dawson, "The effects of fatigue on train handling
during speed restrictions," Transportation Research Part F: Traffic Psychology and Behaviour, vol.
9, pp. 243-257, 2006.
[9] S. K. L. Lal, A. Craig, P. Boord, L. Kirkup, and H. Nguyen, "Development of an algorithm for an
EEG-based driver fatigue countermeasure," Journal of Safety Research, vol. 34, pp. 321-328,
2003.
[10] H. Gu, Q. Ji, and Z. Zhu, "Active facial tracking for fatigue detection," in Sixth IEEE Workshop on
Applications of Computer Vision (WACV 2002), Orlando, Florida, 2002.
[11] W. W. Wierwille, L. A. Ellsworth, S. S. Wreggit, R. J. Fairbanks, and C. L. Kim, "Research on
vehicle-based driver status/performance monitoring: development, validation, and refinement of
algorithms for detection of driver drowsiness," National Highway Traffic Safety Administration Final
Report: DOT HS 808 247 1994.
[12] P. Artaud, S. Planque, C. Lavergne, H. Cara, P. de Lepine, C. Tarriere, and B. Gueguen, "An on-
board system for detecting lapses of alertness in car driving," in 14th E.S.V. Conference, Session 2
- Intelligent Vehicle Highway System and Human Factors, Munich, Germany, 1994.
[13] S. K. L. Lal and A. Craig, "Reproducibility of the spectral components of the electroencephalogram
during driver fatigue," International Journal of Psychophysiology, vol. 55, pp. 137-143, 2005.
[14] T. Gasser, P. Bacher, and H. Steinberg, "Test-retest reliability of spectral parameters of the EEG,"
Electroencephalography and Clinical Neurophysiology, vol. 60, pp. 312-319, 1985.
[15] A. J. Tomarken, R. J. Davidson, R. E. Wheeler, and L. Kinney, "Psychometric properties of resting
anterior EEG asymmetry: temporal stability and internal consistency," Psychophysiology, vol. 29,
pp. 576– 592, 1992.
[16] V. E. Pollock, L. S. Schneider, and S. A. Lyness, "Reliability of topographic quantitative EEG
amplitude in healthy late-middle-aged and elderly subjects," Electroencephalography and Clinical
Neurophysiology, vol. 79, pp. 20-26, 2002.
[17] T. Åkerstedt, G. Kecklund, and A. Knutsson, "Manifest Sleepiness and the Spectral Content of the
EEG during Shift Work," Sleep, vol. 14, pp. 221-225, 1991.
[18] T. Åkerstedt and M. Gillberg, "Subjective and objective sleepiness in the active individual,"
International Journal of Neuroscience, vol. 52, pp. 29-37, 1990.
[19] J. M. Stern and J. Engel, Altas of EEG Patterns. USA: Lippincott Williams & Wilkins, 2005.
[20] L. Torsvall and T. Åkerstedt, "Sleepiness on the job: continuously measured EEG changes in train
drivers," Electroencephalography and Clinical Neurophysiology, vol. 66, pp. 502-511, 1987.
[21] H. J. Eoh, M. K. Chung, and S.-H. Kim, "Electroencephalographic study of drowsiness in simulated
driving with sleep deprivation," International Journal of Industrial Ergonomics, vol. 35, pp. 307-320,
2005.
[22] H. Tietze, "Stages of wakefulness during long duration driving reflected in alpha related events in
the EEG," in Proceedings of the International Conference on Traffic and Transport Psychology
ICTTP, Bern, Zwitserland, 2000.
[23] K. A. Brookhuis and D. Waard, "The use of psychophysiology to assess driver status,"
Ergonomics, vol. 39, pp. 1099-1110, 1993.
[24] T.-P. Jung, S. Makeig, M. Stensmo, and T. J. Sejnowski, "Estimating alertness from the EEG power
spectrum," IEEE Transactions on Bioimedical Engineering, vol. 44, pp. 60-69, 1997.
[25] A. Subasi, "Automatic recognition of alertness level from EEG by using neural network and wavelet
coefficients," Expert Systems with Applications, vol. 28, pp. 701-711, 2005.
[26] M. K. Kiymik, M. Akin, and A. Subasi, "Automatic recognition of alertness level by using wavelet
transform and artificial neural network," Journal of Neuroscience Methods, vol. 139, pp. 231-240,
2004.
[27] C.-T. Lin, R.-C. Wu, S.-F. Liang, W.-H. Chao, Y.-J. Chen, and T.-P. Jung, "EEG-based drowsiness
estimation for safety driving using independent component analysis," IEEE Transactions on
Circuits and Systems I: Regular Papers, vol. 52, pp. 2726- 2738, 2005.
[28] A. Craig, K. Hancock, and M. Craig, "The Lifestyle Appraisal Questionnaire: A Comprehensive
Assessment of Health and Stress," Psychology and Health, vol. 11, pp. 331-343, 1996.
[29] H. H. Jasper, "Report of the committee on methods of clinical examination in
electroencephalography : 1957," Electroencephalography and Clinical Neurophysiology, vol. 10,
pp. 370-375, 1958.
[30] J. Malmivuo and R. Plonsey, Bioelectromagnetism: Principles and Applications of Bioelectric and
Biomagnetic Fields: Oxford University Press, USA, 1995.
[31] World Health Organization, "BMI classification," 2007.
[32] M. Malik, J. T. Bigger, A. J. Camm, R. E. Kleiger, A. Malliani, A. J. Moss, and P. J. Schwartz, "Heart
rate variability: Standards of measurement, physiological interpretation, and clinical use,"
European Heart Journal, vol. 17, pp. 354-381, 1996.
[33] M. Gillberg, G. Kecklund, and T. Åkerstedt, "Sleepiness and performance of professional drivers in
a truck simulator-comparisons between day and night driving," Journal of Sleep Research, vol. 5,
pp. 12-15, 1996.
[34] M. Härmä, M. Sallinen, R. Ranta, P. Mutanen, and K. Müller, "The effect of an irregular shift system
on sleepiness at work in train drivers and railway traffic controllers," Journal of Sleep Research,
vol. 11, pp. 141-151, 2002.
[35] A. Austin and P. D. Drummond, "Work problems associated with suburban train driving," Applied
Ergonomics, vol. 17, pp. 111-116, 1986.
[36] A. Heitmann, R. Guttkuhn, A. Aguirre, U. Trutschel, and M. Moore-Ede, "Technologies for the
Monitoring and Prevention of Driver Fatigue," in Presented at the Driving Assessment 2001: The
First International Driving Symposium on Human Factors in Driver Assessment, Training and
Vehicle Design conference, 2001.
[37] A. Williamson and T. Chamberlain, "Review of on-road driver fatigue monitoring devices," New
South Wales, Australia: NSW Injury Risk Management Research Centre, University of New South
Wales, 2005.
[38] J. Yao, R. Schmitz, and S. Warren, "A Wearable Standards-Based Point-of-Care System for Home
Use," in international Conference ofthe IEEE EMBS, Cancun, Mexico, 2003.
[39] H. A. Thompson, "Wireless and Internet communications technologies for monitoring and control,"
Control Engineering Practice, vol. 12, pp. 781-791, 2004.
[40] L. L.-Y. Shek and Y.-K. Kwok, "An integrated approach to scatternet traffic management in
Bluetooth ad hoc networks," Computer Networks, vol. 45, pp. 99-118, 2004.
[41] C.-H. Shih, K. Wang, and H.-C. Shih, "An adaptive bluetooth packet selection and scheduling
scheme in interference environments," Computer Communications, vol. 29, pp. 2084-2095, 2006.