Attention of Driver during Simulated Drive

Roman Mou

cek and Vojt

ech Ko

Department of Computer Science and Engineering, New Technologies for the Information Society,

Faculty of Applied Sciences, University of West Bohemia, Univerzitn

ı 8, 306 14 Plze

n, Czech Republic

Keywords:

Neuroinformatics, Electroencephalography, Event Related Potentials, Driver’S Attention, Simulated Drive,

Car Simulator, P3 Component, Daytime, Sleep Deprivation, Peak Latency, Fractional 50% Peak Latency,

Fractional 50% Peak Area.

Abstract:

Attention of drivers is a key factor of road safety. Since inattentive drivers cause a considerable number of

accidents, it is worth to examine the causes and course of driver’s attention even in laboratory conditions during

a simulated drive. This paper deals with the experiment in which the methods of electroencephalography and

event related potentials are used under various conditions to investigate driver’s attention. Eleven participants,

university students, were stimulated with audio signals during monotonous drive in four experimental sessions.

The hypothesis is that the peak latency of the P3 component increases in time as the driver is more tired from

monotonous drive, daytime and sleep deprivation. The background of the used methods, experimental design,

participants, data processing, results and ﬁnal discussion are presented in this paper.

1 INTRODUCTION

Attention of drivers is a key factor of road safety.

Inattentive drivers are dangerous to their surroundings

and cause a considerable number of accidents. How-

ever, decline of attention, especially during long rides,

is natural.

In this paper we focus on inﬂuence of monotonous

drive on driver’s attention during simulated drive.

Moreover, attention of driver can be also inﬂuenced

by daytime and sleep deprivation. This is not inves-

tigated by using common behavioral techniques; the

methods of electroencephalography (EEG) and event

related potentials (ERP) are used. An ERP auditory

experiment is performed during a drive (a car simula-

tor is used) and the subsequent analysis of the changes

in the peak latency of the P3 component is investi-

gated. The hypothesis is that the peak latency of the

P3 component (peak latency represents the level of

driver’s attention) increases in time as the driver is

more tired from monotonous drive, daytime and/or

sleep deprivation. University students as tested sub-

jects participated in the EPR experiment; their results

were analyzed and partially interpreted. However,

a deep analysis that includes e.g. statistical evalua-

tion of the results is still in progress and thus it is

not described in this paper. The paper builds on the

already published experiments and results (Moucek

and Rondik, 2012) and (Moucek and Rericha, 2012)

provided by the authors research group. These exper-

iments are based on the same hypothesis but their de-

sign differs and evolves in time according to previous

experience and knowledge.

The paper is organized as follows. Section 2 gives

a short overview of basic principles of the ERP tech-

nique and assumptions related to P3 amplitude and

P3 latency. It provides readers with essential ideas

that are important for the design of experiments de-

scribed further. Then the experiments dealing with at-

tention of drivers are summarized and extended with

respect to the papers (Moucek and Rondik, 2012)

and (Moucek and Rericha, 2012). The objectives of

the designed experiment are given in Section 3. The

description of experimental design, hardware equip-

ment, software tools, participants, course of experi-

ment, environment and obtained data and metadata is

given in Section 4. Data preprocessing is presented in

Section 5; experimental results extended by the ﬁnal

discussion are provided in Section 6.

2 STATE OF THE ART

This section provides a short description of the ERP

technique, the P3 component and especially the re-

lation of P3 amplitude and P3 latency to attention.

Then a short overview of EEG/ERP experiments deal-

543

Mou

cek R. and Koša

r V..

Attention of Driver during Simulated Drive.

DOI: 10.5220/0004934905430550

In Proceedings of the International Conference on Health Informatics (HEALTHINF-2014), pages 543-550

ISBN: 978-989-758-010-9

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

ing with driver’s attention is presented.

2.1 Event Related Potentials and P3

Component

ERPs were ﬁrst used as an alternative to measure-

ments of the speed and accuracy of motor responses in

paradigms with discrete stimuli and responses. They

have two advantages compared to behavioral meth-

ods: they are useful for determining which stage or

stages of processing are inﬂuenced by a given exper-

imental manipulation (a detailed set of examples is

in (Luck et al., 2000)) and they provide an online mea-

sure of the processing of stimuli even when there is no

behavioral response (Luck, 2005).

The P3 component depends entirely on the task

performed by the subject and is not directly inﬂu-

enced by the physical properties of the stimulus. It

is sensitive to a variety of global factors, such as time

since the last meal, weather, body temperature, and

even day time or the time of year (Luck, 2005). Al-

though thousands of experiments related to the P3

component have been published, we still do not know

exactly what the P3 component really means. The

proposal that the P3 component is related to a process

called context updating seems to be approximately

correct (Luck, 2005).

On the other hand, there are known the factors

which inﬂuence the amplitude and the latency of the

P3 component that is sensitive to the probability of

the target stimulus. Ideas and assumptions related to

the latency of the P3 component are associated with

stimulus categorization. If stimulus categorization is

postponed (it also includes increasing the time re-

quired for low-level sensory processing), P3 latency

is increased. While P3 latency depends on the time

required to stimulus categorization, it does not de-

pend on consequent processes (e.g. response selec-

tion). Thus P3 latency can be used to determine if

a performed experiment inﬂuences the processes of

stimulus categorization or processes related to a re-

sponse (Luck, 2005).

In our case we suppose that stimulus categoriza-

tion is inﬂuenced by driver fatigue and the time re-

quired for low-level sensory processing of incoming

stimuli increases with the level of fatigue.

2.2 Experiments on Driver’s Attention

Omitting behavioral studies not many experiments

dealing with driver’s attention during simulated drive

were performed using the techniques of electroen-

cephalography and event related potentials.

Suitability of EEG-based techniques is described

in (Schier, 2000); drivers’ activity during a driving

simulation task was recorded. As the result, an in-

crease in alpha activity was interpreted as less atten-

tional activity and a decrease as more attentional ac-

tivity.

EEG data as an effective indicator to evaluate

driver fatigue are presented in (Li et al., 2012). The

evaluation model for driver fatigue was established

with the regression equation based on the EEG data

from two signiﬁcant electrodes Fp1 and O1. The ac-

curacy of the model was about 92.3%.

The impact of a surrogate Forward Collision

Warning System and its reliability according to the

driver’s attentional state by recording both behav-

ioral and electrophysiological data was presented

in (Bueno et al., 2012). These results showed that

electrophysiological data could be a valuable tool to

complement behavioral data and to have a better un-

derstanding of how these systems impact the driver.

The effect of a normal night’s sleep vs. prior

sleep restricted to ﬁve hours, in a counterbalanced de-

sign, on prolonged (two hours) afternoon simulated

driving in 20 younger and 19 older healthy men was

studied in (Filtness et al., 2012). After sleep re-

striction younger drivers showed signiﬁcantly more

sleepiness-related deviations and greater 4 to 11 Hz

EEG power, indicative of sleepiness.

The ERP technique was used in (Wester et al.,

2008) where the impact of secondary task perfor-

mance (an auditory oddball task) on a primary driv-

ing task (lane keeping) was investigated. The study

showed that when performing a simple secondary task

during driving, performance of the driving task and

this secondary task are both unaffected (Wester et al.,

2008).

3 OBJECTIVES OF

EXPERIMENT

The assumptions described in Section 2 have been

taken into account during designing the experiment.

Then the objectives of the experiment are:

• To construct a monotonous track where a substan-

tial decrease of attention is supposed.

• To design and implement a simple auditory ERP

experiment.

• To perform the auditory ERP experiment on the

group of participants in the following way: each

participant undergoes four drives in a car simula-

tor, these drives are held in two days, one drive

in the morning (between 9 and 12 AM) and one

HEALTHINF2014-InternationalConferenceonHealthInformatics

544

drive in the afternoon (between 1 and 4 PM). The

participant completes ﬁrst two drives after a usual

night’s sleep, while the other two drives are com-

pleted after a sleep restricted to a maximum of

four hours.

• To divide each drive into time intervals of the

same length.

• To compare the latency of the averaged P3 com-

ponents depending on daytime and sleep depriva-

tion and to evaluate results to conﬁrm/reject the

hypothesis given in Section 1.

4 DESIGN OF EXPERIMENT

A simple auditory ERP protocol was designed. All

participants were elicited by the following three audi-

tory stimuli:

• non-target stimulus S1 is a harmonious tone at

a frequency of 560 Hz, duration time 500 ms with

probability of occurrence p = 0.86,

• target stimulus S2 is a harmonious tone at a fre-

quency of 880 Hz, duration time 500 ms with

probability of occurrence p = 0.11,

• rare stimulus S3 is a continuous change of a har-

monious tone at a frequency of 200 Hz to a har-

monious tone at a frequency of 1000 Hz and back,

duration time 1000 ms, probability of occurrence

p = 0.03.

The stimulus onset asynchrony (SOA) was set to

1500 ms. Two target stimuli cannot be sequential.

The stimuli were played from the speakers inside the

car simulator.

Each drive was 20 minutes long and consisted of

four sub-sessions. Each sub-session was ﬁve minutes

long, participants were stimulated in the second and

fourth sub-session. The ﬁrst and third sub-sessions

served both for relaxation of the participant and for

preventing the participant from familiarity with the

presented stimuli. During all stimulation sessions the

participants drove the car simulator on a monotonous

track. The participant responded to the target stimulus

by changing lanes.

4.1 Hardware Equipment

All experiments were performed in the neuroinfor-

matics laboratory at Department of Computer Sci-

ence and Engineering, University of West Bohemia,

equipped with all necessary hardware devices for

EEG/ERPs recording. The experimental car simula-

tor was equipped with the Logitech G27 wheel, ac-

celerator, and brake. Four computers were used: the

ﬁrst one for presentation of stimuli, the second one for

storing recorded data, the third one for the presenta-

tion of the track, and the fourth one for storing video

recordings of drivers from the cab of the car simula-

tor. The track was projected on the wall in front of the

car simulator. V-Amp was used as an EEG ampliﬁer.

4.2 Software Tools

The stimulation protocol was implemented in the Pre-

sentation software tool produced by Neurobehavioral

Systems, Inc (Neuro Behavioral Systems, 2013). The

protocol itself was ﬁve minute long. The sequence

of stimuli was generated randomly, but it always con-

tained the same number of target, non-target, and rare

stimuli. The harmonious tones were generated in the

Audacity software tool. The track was prepared us-

ing the World Racing 2 game produced by the Sy-

netic Company (SYNETIC GmbH, 2013). There was

used the same track as in (Moucek and Rericha,

2012). The BrainVision Recorder (Brain Products,

2013b) was used for recording and storing EEG/ERP

data and the BrainVision Analyzer (Brain Products,

2013a) was used for processing raw EEG/ERP data.

4.3 Recording System

Common EEG caps (the 10-20 system deﬁning the

location of scalp electrodes) were used depending on

the size of the participants’ heads. The reference elec-

trode was placed above the nose and the ground elec-

trodes were placed on ears.

4.4 Participants

A group of 11 volunteers, university students (eight

men, three women), aged 19-23, participated in the

experiment. Table 1 summarizes detailed information

about the participants obtained from completed ques-

tionnaires.

4.5 Course of Experiment

All participants got all necessary information about

the experiment in a written form in advance. Then

informed consent was obtained from all participants.

Before starting each experiment the participant was

familiarized with basic behavioral rules during an

EEG/ERP experiment. Then the participant was fa-

miliarized with the car simulator controls and with

the track, subsequently they were allowed to drive

around.

During the experiment the examiner controlled

data recording, video recording and activated/deacti-

AttentionofDriverduringSimulatedDrive

545

Table 1: Data about participants (Vision - number of diopters, Hours of sleep - usual sleep/sleep before experiment/sleep

deprivation before experiment).

Participant Gender Laterality Vision Hearing Driving license Active driver Hours of Sleep

1 M R - - Y Y 7 / 6 / 4

2 M R 1.75 - Y Y 7 / 8 / 0

3 W R - - Y Y 7 / 7 / 0

4 M R 5 - Y Y 8 / 5 / 2

5 W R - - N N 8 / 8 / 2

6 M R - - Y Y 7 / 7 / 0

7 M R - - Y Y 7 / 7 / 4

8 M L - - Y Y 7 / 10 / 4

9 M L - - Y Y 8 / 8 / 3

10 W R - - Y N 6 / 10 / 2

11 M R 2 - Y Y 7 / 7 / 4

vated the stimulation program. The experiment was

ended after 20 minutes. The participant left the car

stimulator and the examiner asked him/her to ﬁll in

the questionnaire containing the questions related to

his/her feeling of fatigue during/after the ride.

4.6 Environment

All experiments were performed during February 14

and April 12, 2013 (late winter and spring in the

Czech Republic).

4.7 Data and Metadata

EEG/ERP data were recorded with the sampling fre-

quency 1 kHz; no ﬁlters were used during data record-

ing. The resulting signal was stored into three ﬁles:

.eeg ﬁle containing raw data; .vhdr ﬁle containing

metadata that describe raw data in .eeg ﬁle, and

.avg ﬁle containing the averaged signal around the

used stimuli. All recorded data, collected metadata,

and questionnaires were stored in the EEG/ERP por-

tal (EEG/ERP Portal, 2013). The data are publicly

available for registered users.

5 DATA PROCESSING

The recorded EEG/ERP data were further processed

using the following workﬂow:

• Data Filtering: IIR ﬁlter was applied to data from

the Fz, Cz and Pz electrodes. These three elec-

trodes were also selected for further processing.

• Data Segmentation: The epochs were extracted

from datasets, data corresponding to each tar-

get stimulus were selected in the time interval (-

100ms, 900ms) in the area of occurrence of the

target stimulus.

• Rejection of Corrupted Data: The segmented data

containing artifacts were manually rejected.

• Baseline Correction: The baseline was corrected

using the interval (-100ms, 0ms) before occur-

rence of each target stimulus.

• Data Averaging: The epochs selected from each

twenty minutes long experiment of each partic-

ipant were averaged and stored Then the grand

averages for each of four experimental sessions

(morning + usual sleep, afternoon + usual sleep,

morning + sleep deprivation, afternoon + sleep

deprivation) and for some of their combinations

were computed.

The latency of the P3 component was determined us-

ing the following techniques:

• Peak Latency: The simplest way to determine the

latency of the P3 component is to ﬁnd its max-

imum amplitude in the time interval of possible

occurrence of the P3 component. This maximum

value is referred to as peak latency. However,

this measure of peak latency has several short-

comings and is more suitable for the components

with a clearly identiﬁable maximum value. There-

fore, two other techniques of latency determina-

tion were used in this study (Luck, 2005), (Kiesel

et al., 2008).

• Fractional 50% Peak Latency: This technique

simply marks the time point, when 50% percent-

age of the maximum amplitude (this maximum

amplitude is not necessarily the true peak ampli-

tude) was reached in the backward direction.

• Fractional 50% Area Latency: This technique

works by computing the area under the compo-

nent over a given latency range and ﬁnding the

time point that divides that area into halves.

HEALTHINF2014-InternationalConferenceonHealthInformatics

546

−100 0 100 200 300 400

500 600

700 800 900

−15

−10

−5

Time [ms]

Voltage [µV]

morning + usual sleep

afternoon + usual sleep

morning + sleep deprivation

afternoon + sleep deprivation

Figure 1: Grand average on the electrode Fz - experimental sessions differ in the daytime and duration of sleep.

−100 0 100 200 300 400

500 600

700 800 900

−15

−10

−5

Time [ms]

Voltage [µV]

usual sleep

sleep deprivation

Figure 2: Grand average on the electrode Fz after usual sleep and sleep deprivation.

6 RESULTS AND DISCUSSION

The results from performed experiments are summa-

rized in the ﬁgures and tables presented further. The

following description is then valid for each ﬁgure and

table:

AttentionofDriverduringSimulatedDrive

547

Table 2: Latency of the P3 component on the electrode Fz; the peak latency technique applied.

Participant Experimental

session 1 [ms]

Experimental

session 2 [ms]

Experimental

session 3 [ms]

Experimental

session 4 [ms]

1 299 271 276 291

2 300 319 316 320

3 296 298 296 289

4 288 278 280 281

5 286 292 290 284

6 294 301 303 303

7 291 284 300 301

8 284 294 304 298

9 291 299 323 315

10 280 281 279 273

11 291 288 289 296

Grand average GA1 GA2 GA3 GA4

293 295 304 299

Grand average GA12 GA34 GA13 GA24

293 298 293 298

Table 3: Latency of the P3 component on the electrode Fz; the fractional 50% peak latency technique applied.

Participant Experimental

session 1 [ms]

Experimental

session 2 [ms]

Experimental

session 3 [ms]

Experimental

session 4 [ms]

1 333 270 310 323

2 290 283 312 314

3 278 278 272 280

4 270 271 278 263

5 325 291 268 323

6 291 284 289 284

7 269 266 273 269

8 283 265 280 283

9 300 337 320 332

10 269 271 277 268

11 312 314 316 309

Grand average GA1 GA2 GA3 GA4

286 277 284 282

Grand average GA12 GA34 GA13 GA24

281 283 284 279

Table 4: Latency of the P3 component on the electrode Fz; the fractional 50% area latency technique applied.

Participant Experimental

session 1 [ms]

Experimental

session 2 [ms]

Experimental

session 3 [ms]

Experimental

session 4 [ms]

1 378 363 359 375

2 342 343 361 360

3 336 342 346 353

4 329 324 345 320

5 352 351 321 338

6 341 340 345 351

7 319 318 320 317

8 402 390 339 398

9 328 343 354 361

10 337 330 338 334

11 350 337 347 346

Grand average GA1 GA2 GA3 GA4

336 333 344 346

Grand average GA12 GA34 GA13 GA24

335 345 339 339

HEALTHINF2014-InternationalConferenceonHealthInformatics

548

• Experimental Session 1 - a set of experiments per-

formed in the morning after usual sleep

• Experimental Session 2 - a set of experiments per-

formed in the afternoon after usual sleep

• Experimental Session 3 - a set of experiments per-

formed in the morning after sleep deprivation

• Experimental Session 4 - a set of experiments per-

formed in the afternoon after sleep deprivation

• GAX - the grand average computed by averaging

together the averaged waveforms of the partici-

pants of the experimental session X

• GAXY - the grand average computed by averag-

ing together the averaged waveforms of the par-

ticipants of the experimental sessions X and Y

Figure 1 shows the grand averages on the elec-

trode Fz for all experimental sessions and Figure 2

shows the grand averages on the electrode Fz for ses-

sions after usual sleep and sleep deprivation.

Table 2 presents the latency of the P3 component

on the electrode Fz using the technique of peak la-

tency. The resulting data were determined manually

using Brain Vision Analyzer software. Table 3 and

Table 4 show the latency of the P3 component af-

ter application of the fractional 50% peak latency and

fractional 50% area latency. These latencies were de-

termined by using a custom software application.

It is not possible to mutually compare the laten-

cies obtained by using different techniques. The im-

portance of latency is not in its absolute value but in

the difference between the values measured for the

experimental sessions described above.

The results did not demonstrate that the latency of

the P3 component was inﬂuenced by daytime. The

assumption that the latency of the P3 component in-

creases with sleep deprivation can be shown on the

results from Table 2 and Table 4. However, latencies

of individual participants do not conﬁrm this hypothe-

sis when any technique described in this paper is used.

One possible reason for this result is that the number

of the target stimuli for each participant is too small to

get rid of signal noise. This noise is more eliminated

when grand averages are computed.

7 CONCLUSIONS

This paper shortly described the experiment that had

investigated attention of drivers by using the methods

of electroencephalography and event related poten-

tials. Experimental results showed that the P3 com-

ponent had been clearly identiﬁed during all exper-

imental sessions. Despite expectations, prolongation

of peak latency in time was not clearly observed when

the grand average measure of each participant was in-

vestigated. On the other hand, this prolongation was

observable when the techniques of peak latency and

fractional 50% area latency were applied to compute

the grand average for each experimental session. The

results are currently not statistically evaluated to pro-

vide more detailed information.

In the future, it would be probably appropriate to

increase the number of target stimuli (i.e. to prolong

the drive) to get more evident results from individual

participants.

ACKNOWLEDGEMENTS

The work was supported by the UWB grant SGS-

2013-039 Methods and Applications of Bio- and

Medical Informatics and by the European Regional

Development Fund (ERDF), Project ”NTIS - New

Technologies for Information Society”, European

Centre of Excellence, CZ.1.05/1.1.00/02.0090.

REFERENCES

Brain Products (2013a). Brain vision analyzer.

Brain Products (2013b). Brain vision recorder.

Bueno, M., Fabrigoule, C., Deleurence, P., Ndiaye, D., and

Fort, A. (2012). An electrophysiological study of the

impact of a forward collision warning system in a sim-

ulator driving task. Brain Research, 1470:69–79. cited

By (since 1996)1.

EEG/ERP Portal (2008-2013). EEG/ERP Portal.

Filtness, A., Reyner, L., and Horne, J. (2012). Driver

sleepiness-comparisons between young and older men

during a monotonous afternoon simulated drive. Bio-

logical Psychology, 89(3):580–583.

Kiesel, A., Miller, J., Jolicour, P., and Brisson, B. (2008).

Measurement of erp latency differences: A compari-

son of single-participant and jackknife-based scoring

methods. Psychophysiology, 45(2):250–274.

Li, W., He, Q.-C., Fan, X.-M., and Fei, Z.-M. (2012). Eval-

uation of driver fatigue on two channels of eeg data.

Neuroscience Letters, 506(2):235–239.

Luck, S., Woodman, G., and Vogel, E. (2000). Event-related

potential studies of attention. Trends in Cognitive Sci-

ences, 4(11).

Luck, S. J. (2005). An Introduction to the Event-Related Po-

tential Technique (Cognitive Neuroscience). A Brad-

ford Book, 1 edition.

Moucek, R. and Rericha, J. (2012). Driver’s attention dur-

ing monotonous driving. 2012 5th International Con-

ference on Biomedical Engineering and Informatics,

BMEI 2012, pages 486–490.

AttentionofDriverduringSimulatedDrive

549

Moucek, R. and Rondik, T. (2012). Inﬂuence of mental

load on driver’s attention. Transaction on Transport

Sciences, 5(1):21–26.

Neuro Behavioral Systems (2013). Presentation.

Schier, M. (2000). Changes in eeg alpha power during simu-

lated driving: A demonstration. International Journal

of Psychophysiology, 37(2):155–162.

SYNETIC GmbH (2013). World racing 2.

Wester, A., Bcker, K., Volkerts, E., Verster, J., and Ken-

emans, J. (2008). Event-related potentials and sec-

ondary task performance during simulated driving.

Accident Analysis and Prevention, 40(1):1–7.

HEALTHINF2014-InternationalConferenceonHealthInformatics

550