34th Annual International Conference of the IEEE EMBS
San Diego, California USA, 28 August - 1 September, 2012
Abstract— Driving tasks are vulnerable to the effects of sleep
deprivation and mental fatigue, diminishing driver’s ability to
respond effectively to unusual or emergent situations.
Physiological and brain activity analysis could help to
understand how to provide useful feedback and alert signals to
the drivers for avoiding car accidents. In this study we analyze
the insurgence of mental fatigue or drowsiness during car
driving in a simulated environment by using high resolution
EEG techniques as well as neurophysiologic variables such as
heart rate (HR) and eye blinks rate (EBR). Results suggest that
it is possible to introduce a EEG-based cerebral workload index
that it is sensitive to the mental efforts of the driver during
drive tasks of different levels of difficulty. Workload index was
based on the estimation of increase of EEG power spectra in the
theta band over prefrontal areas and the simultaneous decrease
of EEG power spectra over parietal areas in alpha band during
difficult drive conditions. Such index could be used in a future
to assess on-line the mental state of the driver during the drive
It is largely know that driving a car requires a substantial
cognitive effort and attention from driver’s brain. According
to the World Health Organization (WHO) the primary cause
of death in adults from 18 to 29 years old, and the ninth
cause of human death globally, is represented by car
accidents (Preventing Road Traffic Injury: A Public Health
Perspective For Europe, 2009). In fact, we all make
mistakes, even when performing common everyday tasks we
are used to. Depending on the contingent conditions in which
the driver acts, errors can have a significant impact on the
success of the performance or even on the safety of people.
A situation like long distance driving of cars, busses or
trucks in the highway requires attentive and cognitive
resources that could be not at full disposal of the driver at the
end of a working day. Hence, mental workload, fatigue and
drowsiness assessments are really important for improving
the road safety and, consequently, reducing car accidents. In
fact, it is largely known that one of the major sources of
road accidents is due to the drivers’ drowsiness, with the
relative cohort of lapses of attention during the driving. The
analysis of EEG waveforms, and their decomposition in
different frequency bands, have been often employed in the
assessment of the variation of the internal state of the
G. Borghini, C. Caltagirone, J. Toppi, L. Astolfi, R. Isabella and A.
Maglione are with the IRCCS Fondazione Santa Lucia, via Ardeatina 306,
Rome, Italy. D. Wei, W. Kong, Z. Zhou are with Dept. Computer Science,
Hangzhou Dianzi University, Hangzhou, China.
F. Babiloni, G. Vecchiato, S. Vitiello, L. Polidori are with the Dept.
Physiology and Pharmacology, University of Rome Sapienza, P. le A. Moro
5, 00185, Rome, Italy.
(corresponding author e-mail:
subjects during the execution of simple cognitive or sensory-
motor task and it has been demonstrated by several studies
that EEG is sensitive to fluctuations in vigilance and has
been shown to predict performance degradation due to
sustained mental work. The most prominent event is the
increase of the EEG power spectrum in the theta frequency
band over the prefrontal, frontal and parietal cortex, often
located in a midline scalp position . EEG spectral
power in alpha band has been also reported to decrease
during complex and cognitive demanding tasks. Such
decrement occurs over different scalp areas such as the
fronto - central and the parietal ones . These results
suggested to define a workload index by using the variation
of theta and alpha synchronization and, therefore, of their
Power Spectral Density (PSD). In addition, we used also
other neurophysiologic signs of mental activity and
engagement during drive, such as the measurement of the
heart rate (HR) and the eye blink rate (EBR). Both of these
variables have been demonstrated to be correlated with the
mental engage of the driver during the performance. In
particular, it has been suggested that increased HR could be
related with an increased mental workload while eye blinks
(namely duration and frequency) are inversely correlated
with the increase of the mental workload of the drivers .
II. MATERIALS AND METHODS
The subjects were selected by their ages, driving
experiences (i.e. possession of their driving license), cars
that normally they used to drive (i.e. manual gear and not
automatic one) and physical conditions. All the subjects were
prohibited to drink alcohol and to have heavy meals for one
day prior to the measurements, and they were asked to avoid
caffeine, tea or chocolate consumption 5 hours before the
experiment. They were volunteers and experimental
instructions were provided at the beginning of the
B. Experimental protocol
The experiments were performed between 2 PM and 5
PM because day time sleepiness tends to increase during
those hours . The experimental protocol is developed
along 2 days. The selected track was the Spa –
Francorchamps (Belgium) and the car by which performing
the driving tasks was the Alfa Romeo – Giulietta QV (1750
TBi, 4 cylinders, 235 HP). The first day was for training the
subject with the driving simulator and for familiarize with
the task of alert and vigilance (TAV). The alert stimuli, a
white “X”, were presented on a monitor placed 70 (cm) from
the subject just a little below the frontal direction, avoiding
Assessment of mental fatigue during car driving by using high
resolution EEG activity and neurophysiologic indices
G. Borghini, G. Vecchiato, J. Toppi, L. Astolfi, , A. Maglione, R. Isabella, C. Caltagirone, W. Kong,
D. Wei, Z. Zhou, L. Polidori, S. Vitiello and F. Babiloni
6442 978-1-4577-1787-1/12/$26.00 ©2012 IEEE
the interference with the main screen. The vigilance stimuli
were presented by two speakers placed on the left and on the
right side of the driver (Fig. 1). On the second day the
subject had to perform the selected track under different
conditions with the priority of the vehicle control; any
condition was a race of 2 laps. In the first condition
(TASK_0 – day time) the subjects did not receive any
requirements and they were told to do the laps normally, as
the prior day. From these laps, a baseline time was acquired
and in the second condition (TASK_1 – day time) the driver
had to perform the race reducing of the 2% the baseline total
time. The TASK_2’s requirements were like those of the
TASK_1 but performing at the same time the TAV. The
subject could reply to the TAV stimuli by pressing the button
placed on the sides of the steering wheel; the button number
1 (left side) for the vigilance stimuli and the button number 2
(right side) for the alert stimuli. This condition was for
enhancing the workload; the alternating tone sequence of the
vigilance task simulated the car’s radio and the target tone
sequence was a phone calling that the driver had to answer as
quickly as he felt it is safe to do so. The alert’s stimuli
simulated the traffic jam, for example, the traffic lights, the
pedestrians, other cars or other uncontrollable and
unpredictable traffic agents . The fourth and fifth
conditions, TASK_3 and TASK_4, were equal to the
TASK_1 and TASK_2, respectively, but in the night time.
The last condition was a monotonous night driving, in which
the subject had to drive, on the same track, very carefully
without exceeding the speed of 70 (Km/h).
Giulietta QV on the Spa – Francorchamp (Belgium) trak under different
conditions. The different experimental conditions have been set up for
ehnancing the workload, by performing the TAV, and for promoting
drowsiness, by asking the subject to drive slowly with a speed limit.
The experimental protocol consisted in driving the Alfa Romeo
This monotonous task was for making the driver drowsy.
Since the previous task, the drivers used to drive almost fast
and for this reason the requirement of the speed limit was
monotonous enough to
hypovigilance. At the end of each condition the subject had
to fill the NASA-TLX questionnaire for the subjective
workload assessment. The TAV performance and the
telemetry were evaluated by analyzing the log files in which
reaction times, correctness of answers and driving
parameters were recorded automatically.
promote drowsiness or
C. EEG and physiological recording
Electroencephalogram (EEG) and physiological signals,
eye blinks and electrocardiogram (EKG), were recorded by a
digital ambulatory monitoring system (Brain Products
GmbH, Germany). Sixty-one channels EEG and one EKG
channel were collected simultaneously during the experiment
with a sampling frequency of 200 (Hz). All the electrodes
were referenced to both the earlobes and the impedances
were maintained around 10 (kΩ). Electrode for the heart
activity was positioned on the Erb’s point. A 50-Hz notch
filter was applied to all measurements for removing power
interference. The EEG recordings was also band-pass filtered
(low-pass filter cut-off frequency: 40 (Hz), high-pass filter
cut-off frequency: 1 (Hz)) and then the Independent
Component Analysis (ICA) was used in order to remove the
artifacts from the data. By the separation of the independent
sources, the eye blinks component was analyzed for the
estimation of the eye blinks rate (EBR), while the EKG raw
signal was used for estimate the heart rate (HR).
D. High resolution EEG cortical estimation
Cortical activity from EEG scalp recordings was
estimated by employing
technologies, including the use of quasi-realistic head model,
estimation of cortical sources by solving the EEG linear
inverse problem with distributed sources, and statistical
parametric mapping of significant increase or decrease of
EEG spectral activity over the cortex [7-12].
the high-resolution EEG
E. Neurophysiologic indexes for characterizing drowsiness
and mental fatigue.
In the following, we used two major cerebral indices to
characterize drowsiness and the occurrence of mental fatigue
in drivers during the driving tasks. For all concerns
drowsiness recently, the detection of EEG alpha spindles,
defined as short burst in the alpha band was suggested as an
objective measure for assessing the driver’s fatigue during
real driving conditions when compared to the normal EEG
spectral analysis . The occurrence of these bursts is also
consistent with driving error events. Before such events
neurophysiologist posed few criteria, according to their
observations, which can characterize the potential driving
errors as follows “at least 10 (sec) before the driving error
event, alpha synchronization bursts from the lower limit of
alpha waves gradually to theta waves” . Alpha bursts
have frequency between the higher portion of the theta band
and the lower portion of alpha bands, 7 – 9 (Hz), with a
gradual increase of amplitude. As to topographical
distribution, these bursts are more dominant in the central
and parietal areas. We then checked the occurrence of such
EEG events before the drowsiness-induced drive errors (Fig.
2). It has been previously noted as EEG power spectra
increase in the theta band could be correlated with the
insurgence of mental fatigue. In this work it has been defined
a workload index (IWL) as the ratio between the theta psd in
the frontal EEG channels (F3, Fz and F4) and the alpha psd
in the parietal channels (P3, Pz and P4) for the left side
(F3/P3), central line (Fz/Pz) and the right side (F4/P4) of the
brain. The EEG bands were defined by the Individual Alpha
Frequency (IAF) .
A. Visual interpretation of the EEG
Fig. 2 shows an example of the occurrence of EEG alpha
burst (circled in red) occurred during the monotonous
driving task (TASK_5) as signal of drowsiness and reduced
vigilance. Subjects after the appearance of such particular
EEG patterns drove off from the correct trajectory lane with
an high statistical occurrence when compared to the drive
errors performed during standard driving conditions
(p<0.05). In agreement with the previous observations, also
in our case the major occurrence of such EEG bursts were
located in subjects over the centro - parietal scalp areas.
monotonous driving task condition (TASK_5) as signal of drowsiness and
reduced vigilance. These bursts are more dominant in the central and
posterior EEG channels.
Alpha bursts (circled in red colour) occurred during the
B. Workload index
Fig.3 shows an example of how the employed workload
index changed during the different experimental conditions,
respect to the rest condition (TASK_0). For example, in the
central brain region and in the alpha2 band, the conditions in
which the TAV was performed (TASK_2 and TASK_4)
have the highest index’s values and the monotonous task
(TASK_5) the lowest. This was true for almost all sub-
bands, especially in the central and left brain regions.
C. Heart and eye blinks rate
A significant decrease of heart rate (HR) and an increase
of eye blinks rate (EBR) have been observed in the
comparison between all the driving conditions and the
monotonous drive condition (TASK_5) in all the subjects
studied (p<0.05, corrected for multiple comparisons). Fig. 4
shows an example in a particular subject of how the HR
increases (blue line) during the occurrence of high workload
conditions, passing from a quite driving (TASK_0) to
situations in which there are specific requirements, for
example improving the driving performance (TASK_2) or
doing the TAV in the night time (TASK_4). The lower
magenta line shows how the EBR inversely correlates with
difficulties of the driving tasks. In fact, the lower the
difficulty of the task, the greater is the value of EBR; in fact,
the highest EBR’s value was obtained during the
monotonous driving condition and the lowest during the
execution of the TAV in the night time.
faced during the different experimental conditions. As the difficulty of the
task increase, the higher was the IWL’s value. Note that as during the TAV
conditions (TASK_2 and TASK_4) the index has highest values and how
during the monotonous task (TASK_5) the index has the lowest one.
The workload index reflects the mental engagment for the task
D. Estimation of cortical activity
In (Fig. 5) are shown examples of cortical maps relative
to the theta psd differences (statistically significant, p<0.01)
between the rest task (TASK_0) and the task with the TAV
during the day time (TASK_2), on the top of the image and
between the rest task and the monotonous condition
(TASK_5) in the alpha band, on the bottom of the image.
The red colour means that the difference (TASK_2/TASK_5
– TASK_0) is positive, while the blue colour means that the
difference is negative. The red colour means that the psd of
the considered task are greater than those of the rest
condition, thus it is possible to see how with the TAV the
theta psd increases and during the monotonous task the alpha
synchronization increases respect a normal driving condition.
demonstrated that are correlated with the occurrence of the workload; the
HR increases with the workload and it decreases in monotonous and
droswy conditions (TASK_5). On the contrary, EBR is inversely correlated
with the increase of workload.
The HR (blue line) and .the EBR (magenta line) have been
Figure 5. Download full-text
(statistically significant, p<0.01) between the rest task (TASK_0) and the
task with the TAV in the day time (TASK_2), on the top (views a, b, c) and
between the rest task and the monotonous condition (TASK_5) in the alpha
band, on the bottom (views d, e and f). conditions. View (a) is a frontal
perspective, while view (d) is a posterior perspective.
Mean cortical maps reflecting the psd theta differences
signals could help the assessment of the driver’s mental and
physical status, as drowsiness and alertness for the various
driving conditions and how they are correlated with the
driving performance. A common trend has been found; the
higher task’s difficulty, the higher the IWL and HR values,
the NASA-TLX scores, the TAV errors, the mean reaction
times (RT) and the driving errors (off road). In addition, the
EBR decreased with the increase of difficulty and it
decreased during the monotonous condition.
The present study illustrates the development of a
cerebral mental workload index based on EEG data, and the
integration of its information with those of the autonomic
indexes, such as heart rate and the eyeblinks rate. The
proposed workload index could be estimated in a near future
in “real-time” system and could be eventually integrated
inside the car to provide a feedback to the user about its
internal cognitive conditions during the drive task.
The results show how these neurophysiological
This work was partly supported by Major International
Cooperation Project of Zhejiang Province, No.2011C14017.
Thanks to Alexander Tropakov for the World Records
software by which it was possible to save the telemetry data
of any driving tasks (wrecord.com).
 C. Berka, D. J. Levendowsky, M. N. Lumicao, A. Yau, G. Davis, V.
T. Zivkovic, R. E. Olmstead, P. D. Tremoulet, P. L. Craven, “EEG
correlates of task engagement and mental workload in vigilance,
learning, and memory tasks”, Aviat Space Environ Med, vol. 78, no.
5, pp. B231 – B244, 2007.
 G. Borghini, R. Isabella, G. Vecchiato, J. Toppi, L. Astolfi, C.
Caltagirone, F. Babiloni, “Brainshield: HREEG study of perceived
pilot mental workload”, Italian journal of aerospace medicine, vol. 5,
pp. 34 – 47, 2011.
 S. H. Fairclough, L. Tattersall, “The influence of task demand and
learning on the psychophysiological response”, International Journal
of Psychophysiology, vol. 56, no. 2, pp. 171 – 184, 2005.
 S. K. L. Lal, A. Craig, “Driver Fatigue: psychophysiological effects”,
The Fourth International Conference on Fatigue and Transportation,
 S. D. Baulk, L. A. Reyner, J. A. Horne, “Driver Sleepiness –
Evaluation of Reaction Time Measurement as a Secondary Task”,
Sleep, vol. 24, no. 6, pp. 695-698.
 K. M. Bach, M. G. Jæger, M. B. Skov, N. G. Thomassen, “Interacting
with In – Vehicle System: Understanding, Measuring, and Evaluating
Attention”, HCI 2009 – People and Computers XXIII – Celebrating
people and technology, pp. 453 – 462.
 C. Babiloni, F. Babiloni, F. Carducci, F. Cincotti, F. Vecchio, B.
Cola, S. Rossi, C. Miniussi and P.M. Rossini. Functional
frontoparietal connectivity during short-term memory as revealed by
high-resolution EEG coherence analysis. Behavioral Neuroscience,
 L. Astolfi, F. Cincotti, C. Babiloni, F. Carducci, A. Basilisco, P.M.
Rossini, S. Salinari, D. Mattia, S. Cerutti, D. Ben Dayan, L. Ding, Ni
Y, B. He, F. Babiloni. Estimation Of The Cortical Connectivity By
High Resolution EEG And Structural Equation Modeling:
Simulations And Application To Finger Tapping Data, IEEE Trans
Biomed Eng. 2005 May;52(5):757-68,
 C. Babiloni, A. Brancucci, F. Babiloni, P. Capotosto, F. Carducci, F.
Cincotti, L. Arendt-Nielsen, AC. Chen, P.M. Rossini. Anticipatory
cortical responses during the expectancy of a predictable painful
stimulation. A high-resolution electroencephalography study. Eur J
Neurosci. 2003 Sep;18(6):1692-700
 L. Astolfi, F. Cincotti, D. Mattia, S. Salinari, C. Babiloni, A.
Basilisco, P.M. Rossini, L. Ding, Y. Ni, B. He, M.G. Marciani, F.
Babiloni. Estimation of the effective and functional human cortical
connectivity with Structural Equation Modeling and Directed
Transfer Function applied on high resolution EEG, Magn Reson
Imaging. 2004 Dec;22(10):1457-70
 L. Astolfi, F. Cincotti, D. Mattia, M.G. Marciani, L. Baccalà, F. de
Vico Fallani, S. Salinari, M. Ursino, M. Zavaglia, F. Babiloni
Assessing Cortical Functional Connectivity By Partial Directed
Coherence: Simulations And Application To Real Data. IEEE Trans
Biomed Eng. 2006 Sep;53(9):1802-12
 L. Astolfi, F. De Vico Fallani, F. Cincotti, D. Mattia, M.G. Marciani,
S. Bufalari, S. Salinari, A. Colosimo, L. Ding, J.C. Edgar, W. Heller,
G.A. Miller, B. He, F. Babiloni Imaging Functional Brain
Connectivity Patterns From High-Resolution EEG And fMRI Via
Graph Theory, Psychophysology, 44(6):880-93, 2007
 M. Simon, E. A. Schmidt, W. E. Kincses, M. Fritzsche, A. Bruns, C.
Aufmuth, C. Bogdan, W. Rosenstiel, M. Schrauf, “EEG alpha spindle
measures as indicators of driver fatigue under real traffic conditions”,
Clinical Neurophysiology, vol. 122, no. 6, pp. 1168 – 1178, 2011.
 C. Papadelis, Z. Chen, C. Kourtidou-Papadeli, P. D. Bamidis, I.
Chouvarda, E. Bekiaris, N. Maglaveras, “Monitoring sleepiness with
on – board electrophysiological recordings for preventing sleep -
deprived traffic accidents”, Clinical Neurophysiology, vol. 118, pp.
1906 – 1922, 2007.
 W. Klimesch, “EEG alpha and theta oscillations reflect cognitive and
memory performance: a review and analysis”, Brain Research
Reviews, vol. 29, pp. 169-195, 1999