You are on page 1of 14

REVIEW PAPER

International Journal of Occupational Medicine and Environmental Health 2015;28(6):941954


http://dx.doi.org/10.13075/ijomeh.1896.00317

APPLICATION OF EYE-TRACKING IN DRIVERS


TESTING: AREVIEW OF RESEARCH
BRONISAW KAPITANIAK, MARTA WALCZAK, MARCIN KOSOBUDZKI, ZBIGNIEW JWIAK,
andALICJABORTKIEWICZ
Nofer Institute of Occupational Medicine, d, Poland
Department of Work Physiology and Ergonomics

Abstract
Recording and analyzing eye movements provide important elements for understanding thenature of thetask of driving
avehicle. This article reviews theliterature on eye movement strategies employed by drivers of vehicles (vehicle control,
evaluation of thesituation by analyzing essential visual elements, navigation). Special focus was placed on thephenomenon
of conspicuity, theprobability of perceiving an object in thevisual field and thefactors that determine it. Thearticle reports
themethods of oculographic examination, with special emphasis on thenon-invasive technique using corneal reflections,
and thecriteria for optimal selection of thetest apparatus for drivers in experimental conditions (on adriving simulator)
and in real conditions. Particular attention was also paid to thehelmet or glass-type devices provided with1 or2high
definition(HD) camcorders recording thefield of vision and thedirection of gaze, and thenon-contact devices compris-
ing2 or3cameras and an infrared source to record eye and head movements, pupil diameter, eye convergence distance,
duration and frequency of eyelid blinking. Areview of thestudies conducted using driver eye-tracking procedure was pre-
sented. Theresults, in addition to their cognitive value, can be used with success to optimize thestrategy of drivers training.

Key words:
Eye-tracking, Visual strategy, Drivers, Visual fatigue, Road accidents

HISTORICAL BACKGROUND contribution [1]. In 1879, Javal invented an ophthalmo-


The interest in eye movement is not new. As early as in scope used to examine eye movements, by theuse of which
the 11th century, Ibn al Haytam described, for the 1st he described the1stsaccades and fixations. Acomprehen-
time, fast eye movements that occurred when tracking sive historical overview of research on eye movements was
moving objects. However, it was not until the18thcentury presented by Wade in2010[1].
that thestudies on thenature of these movements were The1stinvasive registration of eye movements using con-
undertaken. The1stdescription of these movements was tact lenses(1897) is attributed to2scientistsHuey and
presented by Porterfield(1737) and then, later in1792[1], Delabarre [1]. The 1st non-invasive recordings, made in
Wells described fast movements of eyes, which he called theearly20thcentury, using optical instruments were per-
nystagmus. In the 19 century, Purkynjs studies on formed by many researchers, among them: Dodge, Erd-
themechanisms of perception of brightness were amajor mann, and Cline[1]. Since1905, film cameras have been
This work was prepared partly in the frame of project entitled: Integrated system for monitoring the psychical and physical condition of road vehicle drivers to
minimise hazards in road traffic financed by Innovative Economy Programme: Priority 1. Research and Development of Modern Technologies Activity
(UDA-POIG.01.03.01-10-085/09). Project manager: Prof. Alicja Bortkiewicz.
Received: May 6, 2014. Accepted: March 26, 2015.
Corresponding author: A. Bortkiewicz, Nofer Institute of Occupational Medicine, Department of Work Physiology and Ergonomics, w. Teresy 8, 91-348 d, Poland
(e-mail: alab@imp.lodz.pl).

Nofer Institute of Occupational Medicine, d, Poland 941


REVIEW PAPER B. KAPITANIAK ET AL.

used for that purpose notably by Judd, MacAllister, Bus- information technology and miniaturization of optical
well, andSteel. In1921, Gilliland performed a2-dimen- equipment. A comprehensive review of the literature in
sional recording of eye movements[2]. that area was published in1998 by Rayner who analyzed
Research on visual reading strategy started in the1930s[1]. over900publications[12]. It focused on thestudies of eye
In1931, Earl and Taylor developed2devices, called Oph- movements during reading and other information process-
thalmograph and Metronoscope, that enable accurate ing tasks, such as music note reading, writing, or observing
recording of eye movements, and above all, of saccades ascene. Themain focus of thereview was on reading as
and fixations[1]. aspecific example of cognitive processing.
The1steye-tracker (adevice for recording eye move- The main issues in relation to reading are: characteristics
ments) in aform of ahelmet equipped with amini-camera of eye movements, integration of information through
was the work of Hartridge and Thompson (1948) [3]. It theanalysis of saccades, eye movement control and indi-
was improved by Shackel in 1960 [4], and then in 1962 vidual differences. Theauthor summarizes thereview by
by Mackworth and Thomas[5]. In1967, Yarbus[1] start- claiming that the analysis of eye movements has proved
ed the research using his oculometer, and thus, opened to be extremely valuable in thestudy of theprocesses of
a wide, new field of research. The era of modern eye- perception and information processing. Unfortunately,
tracking began in the 1970s with the studies performed the results cannot be generalized to various forms of
byYoung and Sheena and their successors[6]. cognitive tasks, but characteristics of eye movements in
The use of eye-tracking to control visibility of theelements a variety of tasks are necessary to understand cognitive
of visual advertising was thework of an advertising agency strategies.
EuroRSCG and it started in thelate1970s. Followers of In our paper we will try to present theactual status of re-
these applications included other public agencies and, search, in particular concerning thevisual strategy studies
later on, website creators. This technique made it possible on driving. This presentation is not exhaustive, but it can
to optimize and minimize their advertising messages while be useful for thefuture perspectives of research.
ensuring their better advertising effect.
Application of this technique to observe driving strate- TYPES OF EYE MOVEMENTS
gies also dates back to the1970s. Already in1971, Soliday During a visual task, basic eye movements include sac-
published asurvey of reports on thestudies of drivers[7]. cades rapid eye movements at aspeed of up to500 pers
One of the1st devices used in these studies was theJapa- or even800 pers. Theamplitude does not exceed8 and
nese NACEye Marker Recorder. At that time, studies on theduration is20200ms, depending on theamplitude.
thedifferences in visual strategies of theexperienced and Thelatency thetime from themoment atarget appears
novice drivers started and concept of conspicuity was in the visual field to the start of eye movement rang-
coined, denoting theability to perceive an element, associ- esfrom 100300ms. It depends primarily on thequality
ated with aconcept of thefunctional field of vision[810]. of cognitive process, increasing with its complexity. La-
In 1972 Mourant and Rockwell [11] found different vi- tency is also increased when thefunctional field of view
sual patterns in thenovice and experienced drivers with extends to theperipheral zone, or when ahigh precision is
theTVeye-movement system. required in atarget location.
In the1980s and1990s, extensive research on eye move- The sensitivity of visual perception is much reduced during
ments took place, mainly related to the development of saccades; thephenomenon is called saccadic suppression

942 IJOMEH 2015;28(6)


EYE-TRACKING AND TESTING OF DRIVERS REVIEW PAPER

and it is thesubject of aserious debate. Its effect is not movements are observed vestibulo-ocular movements
conscious, most likely due to theexistence of theinhibit- and optokinetic nystagmus. These movements serve to
ing mechanism in thecentral nervous system[13]. stabilize an image in the foveal zone and, in contrast to
Many studies cited by Rayner suggest that some cognitive thepursuit, are not consciously controlled. Thevestibulo-
functions are inhibited during saccades, at least for simple ocular movements are primarily associated with move-
tasks[12]. However, aperson does not realize thegaps in ments of the head and are generated by the reaction of
mental activity during eye movements because thesaccade thevestibular system in theinner ear. They are very fast
duration is very short. Studies have shown that the pro- (faster than the pursuit) and precise. In contrast, opto-
cesses of lexical analysis while reading are not inhibited kinetic nystagmus is a reflex generated by translations
during saccades[14]. of theoptical system taking place in thecentral nervous
Saccades should be distinguished from 3 other types of system[16,17].
eye movements pursuit, vergence, and vestibular move- Although a lot of research concerns importance of
ments (synchronizing eye movements with movements of these3types of movements, saccade movements are con-
thehead). sidered to be themost important for thecognitive process.
Pursuit occurs when our eyes follow a moving target; It also seems worthwhile to mention3other types of eye
the speed of those eye movements is much slower than movements, characterised by high frequency and low am-
the saccade below 30/s. If a target is moving quickly plitude: nystagmus, drifts and microsaccades. Although
across theentire field of vision, pursuit is often accompa- analysis of eye movements makes use of the concept of
nied by asaccade to catch up with thetarget. Some authors thefixation relative to interrupting eye movement at some
claim that it is desirable to analyze slow eye movements to point between2saccades, in fact theconcept is mislead-
evaluate thestate of adrivers activity[15]. ing. Eyes are never really fixed, because physiological nys-
Vergence eye movements occur when theeyes are guided tagmus is continuous and has aa nature of low-amplitude
to thecentre of thefield of view to focus on anear object. (<1) and high-frequency vibrations(50100Hz). Its role
They reflect thecompensation of binocular vision at vary- is not entirely clear, although it is often assumed that these
ing distances. When atarget of observation is located at oscillations are related to theperceptual activity of neu-
adistance of far vision (>6m), eye movements are paral- rons in theretina. Drifts (movements with an amplitude
lel. They become convergent when thetarget is located at of 2120) and microsaccades (fast movements with an
adistance of near vision (<1m), as they serve to position amplitude of260) are movements theprecise meaning
theimage within thefoveal area (about2 around thevi- of which is not known, but it is assumed that they repre-
sual axis). These movements are also used to track an ob- sent the correction of inaccurate eye movement control
ject when it is approaching or receding. Tracking an object process.
that moves in3dimensions that is moves laterally and Permanent movement of images on the retina in an ir-
recedes at thesame time requires combined pursuit/ver- regular manner constitutes thecumulative effect of these
gence movements. Vergence movements are usually slow, micro-movements. Theamplitude of these movements is
rarely exceeding10/s. about 0.1, while the distance between the receptors in
Vestibular eye movements appear to compensate for move- theretina is about0.01, which is necessary for theper-
ments of thehead and body, in order to maintain thesame ception of theimages. Pharmacological blocking of these
direction of aview. When thehead moves,2types of reflex movements causes functional blindness.

IJOMEH 2015;28(6) 943


REVIEW PAPER B. KAPITANIAK ET AL.

Between the 2 saccades, eyes stop on the object being not allow to track thedirection of gaze, since it can track
viewed for about200600ms, and thestops are referred eye movements only when thehead is immobilized. It is not
to as fixations. Their duration depends mainly on thein- suitable for long-term recording either, due to theneed of
tensity of theluminance contrast, complexity of thearea adhesive fastening of theelectrodes around theeyes. How-
of vision and cognitive load. Depending on thedifferent ever, it has been applied with success in theconditions of
task of reading, fixations have different characteristics; experimental studies using asimulator [22].
their duration varies between 200 ms on silent reading The galvanometric technique consists in measuring chang-
to400ms on typing (Rayner)[12]. es in theelectromagnetic field by means of contact lenses
In thestudies concerned with visual information process- applied onto thecornea, equipped with an induction coil.
ing, as well as with general visual strategy, these are pre- It is very precise; however, due to its invasiveness, it is very
cisely those characteristics that are most commonly used restrictive and impossible to use in thefield studies.
to quantify eye movement. The corneal reflection technique is currently the only
In the task of driving, the saccades could be shorter technique that allows tracking direction of gaze in anon-
than 200 ms [18]. In the study concerning responses to invasive way. It involves using asource of infrared light re-
thecritical hazard events, in adriving simulator, Horrey flected from thecornea, which can be tracked in thefield
and Wickens using the Smart Eye Pro system have ob- of view. This is possible owing to thefact that thereflec-
served thesaccades from60ms to200ms[19]. They have tion from themoving parts of thepupil can be separated
also measured glance duration which is defined astime from thereflection from thestatic elements of thecornea,
of watching an object. This time in their study was0.13s. whereby it is possible to precisely determine theposition
Martens and Fox[20] have performed thesame measure- of eye from therelative positions of thereflections. This
ment of glance duration in a study about recognition of particular technique is currently being used for long-term
traffic signs in driving alow cost simulator. Using theeye- tracking of eye movements in thefield studies[23,24].
trackerISCAN they have found durations from0.3 to2s.
Eye movement recording methods
Oculometry usedinthefieldstudies
Oculometry is amethod that allows measurement of eye The studies of visual strategies during visual tasks are car-
movements and gaze direction tracking in thevisual field ried out mainly by means of theapparatus using thecorne-
test. Historically,3types of oculometry were used: elec- al reflection technique. Therecent years have seen arapid
trooculographic, galvanometric and the corneal reflec development of that technique, which enabled production
tiontechniques. of a number of devices for long-term recording in most
The electrooculographic technique (EOG) involves mea- work situations. It involves superposing thereflection of
surement of bioelectric potential differences caused by cor- an infrared light ray sent in the direction of the cornea
neal-retinal eye movements. These potentials are recorded from asource placed adjacent to theeye, onto theimage
using electrodes placed around theeyes. Thereceived sig- of thefield of view filmed by acamera mounted on drivers
nal is filtered to eliminate artefacts. This technique is very head. Suitable software enables precise tracking of gaze
precise in assessing the range of motion with good time direction.
resolution and poor spatial resolution. It is one of the1st The corneal reflection technique cannot be used in con-
techniques used in clinical trials[21]. This technique does ditions of high-intensity solar radiation, which makes it

944 IJOMEH 2015;28(6)


EYE-TRACKING AND TESTING OF DRIVERS REVIEW PAPER

impossible to identify theinfrared ray produced by theap- Reliability thecamera should work reliably in every situ-
paratus. In this case, you can replace it with acomputer ation and with every person; poor lighting conditions may,
program that simulates thedrivers head and eyes to assay however, interfere with its operation; interferences may
gaze direction, with aslightly lower accuracy than that of also occur in people who wear contact lenses or glasses.
thecorneal reflection method. Thecomputer-simulation Low-level filter the data recorded directly by the de-
method requires ahigh efficiency computer system able to vice are subject to artefacts related to blinking or clos-
produce such amodel in real time, and necessitates very ing theeyelids; filtering and smoothing thesignal causes
high precision in setting thecameras. Currently it is used, thedestruction of information on themovements of small
among others, in theFaceLab system. amplitude; thetype of filter used determines themove-
ments actually taken into account in thefurther analysis.
The criteria used in selecting methods Programming and data visualization interface most
ofregistrationduring thefield studies devices have a software package for visualizing and
Currently in thefield studies,2types of devices are used: evaluating the recorded eye movements; however, it
devices placed on drivers head, such as ahelmet or glass- is very important to be offered thepossibility of using
es, and independently installed devices without physical additional applications that make it possible to adapt
contact with theperson to be tested. Thelatter are par- themachine to thetest better.
ticularly valuable in thefield studies, where theuse of hel- According to Ouldetal., these criteria should be extend-
mets and even glasses is not always possible or does not ed [26]. In the opinion of that author, who provides an
guarantee themaximum safety of thetest. overview of the literature on the subject, a good appa-
Drewes has performed a study about following visual ratus for eye-tracking study should meet thefollowing
targets on screen in order to optimise eye-gaze user in- requirements[26]:
terface design[25]. After adeep review of current litera- allow access to thefield of view without distortion, with
ture he has proposed criteria for optimal eye movement an easy access to theface and head,
recording apparatus concerning eye cameras as well as operate without any contact with thesubject,
scene cameras. Each criterion is important, depending artificially stabilize theretinal image as necessary,
on theapplication of thedevice and therange of tests to be accurate up to at least afew minutes of arc, theaccu-
be performed using that device[25]. Here are the5crite- racy is limited by thecombined effects of non-linearity,
ria he distinguishes: distortion, delay and other sources of error,
Accuracy thelimit of accuracy of theeye is approxi- have aresolution of1arcminpers, and thus, be able to
mately 0.5; most devices meet this criterion; only detect theslightest changes in theposition of eye,
studies on micro-saccades and rapid eye movements have a wide dynamic range of movement: from 1min arc
require better accuracy. to 45 for theposition of eye and1800/s for thespeed
Resolution each device records gaze direction at of eye movement,
specified intervals. The video camera recording fre- have agood rate of reaction (e.g.,100Hz),
quency is usually determined by thespecified number be able to give an answer in real time,
of frames per second, which is not always sufficiently be able to make measurements in3planes of eye move-
high for a precise observation of saccades, especially ments (anterior-posterior, sagittal and transverse) and
inhuman/computer interface studies. be insensitive to thetranslation of eye,

IJOMEH 2015;28(6) 945


REVIEW PAPER B. KAPITANIAK ET AL.

be easily adaptable to binocular recording, independent of a device that could potentially limit his
not interfere with theregistration of other physiological freedom of movement, or even visibility.
functions, Adevice of this type consists of2or moreHDcameras
be easy to use by different subjects. and an infrared source placed between the cameras.
The cameras are placed in front of the face of the sub-
Apparatus for testing thevisual strategy ject in aspecific location relative to thesubjects eyes. An
There are3main types of apparatus for testing visual strategy: additional camera can be set so as to record thesubjects
helmet devices, contactless devices and special equipment. field of vision. The device simultaneously captures eye
and head movements, pupil diameter, eye convergence
Helmet devices distance and duration, and frequency of eyelid blinking.
It was the1stdevice used to test visual strategies while driv- Some devices also have theability to track gaze not only
ing. One of thefirst was theNacEye Marker, thenewer with infrared reflection from thecornea, but also by cre
versions were used to test drivers in the1970s and1980s[1]. ating acomputer model of thehead and eyes. This ensures
Currently, the apparatus used for this type of research that thedevice operates effectively even under strong sun-
is extremely miniaturized to allow theuse of lightweight light, which excludes theuse of infrared light. Addition-
glasses instead of afull helmet. ally, there is also theability to analyze facial expressions
The apparatus characterized by high accuracy of themea- based on thedynamics of movements of themouth, eyes,
surements can be designed either for use with an immobi- eyebrows,etc.
lized, or afreely movable head. They are suitable for use Simpler devices are provided for testing visual strategy
in a laboratory, where they can work smoothly together during observation of thescreen. AnHDcamera placed
with apparatus for recording biological signals(electroen- below the screen makes it possible to record eye move-
cephalography (EEG), electromyography (EMG), mag- ments, head movements and pupil diameter. They are
netic resonance imaging(MRI)). used primarily for analyzing tasks of reading thescreen,
Nowadays, the apparatus used for field studies consists and therefore, can be used for research in a simulator,
mostly of lightweight glasses fitted with1 or2high defini- while they are not suitable for real-time test driving. Ow-
tion(HD) digital cameras recording thefield of vision and ing to the integration of the eye-tracking system with
direction of gaze, and aminicomputer. Therecorded data thecomputer screen, theapparatus is very convenient for
can be observed in real time. practical use. Theonly limit is your computer screen that
The new family of devices combines2technologies ultra high restricts thescope of control of eye movements.
speed tracker for recording movements in3D, and power-as-
sisted eye-movement-controlled system for controlling move- Special devices
ments of therecording cameras. Superposing thefield of vision These include equipment used in basic research. In
on thewide-angle panoramic camera image results in ahybrid thelast decade, rapid progress in studies on neurophysi-
image, allowing avery precise tracking of gaze direction. ology has been possible thanks to breakthroughs in im-
aging techniques such as Positron Emission Tomogra-
Contactless devices phy (PET), computed tomography, magnetic resonance
The contactless apparatus is much better suited for test- imaging(MRI). Thus, test devices have been developed,
ing drivers due to the fact that the driver is completely adapted to track eye movements in atunnel, or in other

946 IJOMEH 2015;28(6)


EYE-TRACKING AND TESTING OF DRIVERS REVIEW PAPER

similar conditions. In order to reliably classify the areas Thefield of view may, therefore, be functionally divided
of brain involved in the control of eye movements, it is into 3 regions, in which[28]:
important to accurately measure actual eye movements. the stimuli can be identified without eye movement,
For this purpose, theinfrared ray is guided without con- eye movement is necessary to identify thestimulus,
tact through theoptic fibre to theeye of thesubject. Two identification requires head movement.
additional optic fibre cables receive thelight reflected by
thecornea. By calibrating theeye position in thescanner, CONSPICUITY
it is possible to estimate theposition of an eye with high Another problem is the ability to perceive objects in
temporal and spatial resolution. the visual field. Conspicuity is defined as the prob-
Recent years have seen considerable progress in re- ability of perceiving an object in the visual field [29].
search on the use of catadioptric cameras [27]. These Among the authors there is no agreement on a clear
cameras are made up of a system of mirrors and lenses definition, and many of them point to different aspects
to produce a panoramic view. They have been used for of thephenomenon. Many of them focus primarily on
at least10years in robotics, in particular in theconstruc- thecharacteristics of theviewed objects and theobserv-
tion of self-moving robots. Recently they have been used ers attentiveness. These characteristics vary depending
as a driving aid. A device for recording eye movements on whether the observed object is actively sought or
using this type of camera is currently being developed[27]. unexpected[30].
In neuroscientists approach, conspicuity is considered
Visual strategy on driving as an attentional process, which could be performed
During looking ahead, the field of view may be divided on 2 component frameworks, one a bottom-up, fast,
into 3 zones: foveal, parafoveal and peripheral. Visual primitive mechanism that biases the observer towards
acuity is excellent only in thefoveal zone, which is located selecting stimuli based on their visibility (saliency), and
within2 from thecentral axis of vision. Theacuity is much the 2nd a slower, top-down mechanism with variable
poorer in theparafoveal area (about5 from thecentral selection criteria under cognitive, voluntary control[31].
axis of vision), and quite weak in the peripheral zone. Quantitative evaluation of conspicuity is difficult, espe-
Hence, thetendency of eye movements intended to ensure cially if it is to be objective. The most commonly used
good visual acuity in the whole field of view. Thus, sac- methods include the observers verbal description and
cades serve to quickly include thefield of observation into registration of eye movements.
thefoveal zone, and fixations are necessary to recognise Cole and Jenkins (1980) propose the following quanti-
thedesired information in thefield. tative definition of conspicuity: an object is conspicuous
The results cited by Rayner suggest that the required if90% of theobservers have noted it within0.25s, regard-
precision of recognition is essential for generation of sac- less of its location in thevisual field[32].
cades [12]. It has been observed that, in the parafoveal Difficulties in assessing conspicuity through verbalization
zone, theobjects which are large enough are recognized come mainly from differences in training and theability
without specific eye movements, which points to theexis- level of the observer. These differences may render this
tence of aprocess control system within thecentral ner- type of evaluation quite unreliable.
vous system. This process control system could be con- Registration of eye movements is currently theonly meth-
sidered as a cognitive recognition of visual perception. od that allows for an objective quantitative assessment of

IJOMEH 2015;28(6) 947


REVIEW PAPER B. KAPITANIAK ET AL.

conspicuity[33,34]. Its use in theprocess of driving acar depending on thecontents of thesign. Most authors have
isbecoming more and more widespread. found that when driving in urban conditions, thenumber
of the read signs decreases along with the intensity of
VISUAL TASKS RELATED TO DRIVING AVEHICLE traffic. Some authors propose to perform measurements
We can distinguish3basic visual tasks while driving: of the maximum distance from which a driver can read
control of thevehicle, thesign and thedistance from which he/she actually begins
assessment of thesituation by analyzing essential visual to read it. Theratio of these2distances would be agood
elements, indicator for aproper location of traffic signs[37,38].
navigation. The conspicuity of objects in a drivers field of view de-
During navigation, adriver tends to choose atarget point pends on 2 categories of factors associated on the one
in thefield of view, allowing him/her to anticipate direction hand, with thecharacteristics of thesign (color, size, light-
of travel of thevehicle. This tendency can cause2types ing,etc.), or thecomplexity of thesituation in thefield of
of conflict: emergence of anumber of points that can be view, and on theother hand, with thedriving task. It seems
target points in thesame visual task, or occurrence of that themost important factor determining efficiency of
anumber of elements that could be important for driving, perception is the complexity of the situation in the field
but related with avariety of other visual tasks, as pedes- of view.
trian detection or visual control of navigation system[35]. Quantitative assessment of complexity of the scene is
The most important source of information that makes car extremely difficult, because objective methods are not
driving possible comes from the moving elements which available. Most authors make use of subjective criteria
appear in adrivers field of vision. Observation of those which, however, cannot be used for the quantitative as-
elements provides information that helps to assess thesit- sessment. Characteristics of theobject can be more objec-
uation and ensure smooth navigation. Adriver does not tive, because thesize, colour, contrast or brightness can be
knowingly receive all the information coming through measured[34].
theoptic canal, but only that which is essential for his/her With simple scenes, conspicuity is limited only by thevis-
security. The probability to detect an object while driv- ibility of an object. For complex scenes, the situation is
ing (conspicuity) is very important for theassessment of complicated because of theimportance of cognitive mean-
driving effectiveness. Unfortunately, there is currently ing (sense) of theobjects present in thefield of view. Un-
no objective method for quantitative assessment of that fortunately, it is not always possible to apply theresults of
probability[36]. research carried out in laboratory in controlled situations
The results of questionnaire surveys and eye-tracking tests to theprocess of real driving, mainly due to thefact that
suggest that only about 10% of the written information this process applies to all multiple visual tasks at once,
present in adrivers field of vision is read by thedriver[34]. rather than to asingle task. However, some of theresults
The remaining information is neglected by the driver. have been confirmed by tests in real conditions, such as
Thecorrect choice of information relevant to thesafety of deterioration of perception ability due to thepresence of
driving determines efficiency of adriver. In an experimen- thebillboards in thecomplex field of view on thenational
tal situation when adriver is instructed to read all traffic road[34].
signs, his/her visual strategy changes compared to normal Results of the studies on the visibility of an object are
driving. In this case, reading time ranges 250750 ms, used to determine thecontrast threshold(Ls/L ratio

948 IJOMEH 2015;28(6)


EYE-TRACKING AND TESTING OF DRIVERS REVIEW PAPER

of thedifference of themaximum luminance to minimum that are on theroad and on thepavement. Their strategy is
luminance values) depending on thesize and position of to minimize therisks associated with other road users[39].
an object relative to thecentral axis of vision, and com- Fildes et al. [40] have measured, by means of a non-
plexity of thescene. contact FaceLab, the same parameters, i.e., the number
Complexity indicators that seem to be important in pre- of fixations, average fixation time, percentage of area in
dicting detection of visible elements are based on theas- the total field of view during fixation [40]. This appara-
sessment of the number of objects observed, number of tus allows aprecise decomposition of thefield of view, in
light sources, number of objects in the10cone, and evalu- terms of dividing this field into test areas, as in theEye-
ation of mental workload. Since theaverage reading dis- Works software. Tests were made on asimulator and their
tance of an object is at about2/3of thedistance of maxi- goal was to compare the strategy of young (2535 year
mum readability, it seems logical to conduct evaluation of old) and older (6575 year old) drivers when driving in
complexity at this particular distance. urban conditions and on thehighway. Theresults indicate
It is extremely difficult to assess characteristics of thedriv- that older drivers assess traffic parameters more slowly
ing task objectively, especially in areal situation. Studies (their fixation time is longer than in theyoung) and do it
in laboratory conditions are always subject to amethodol- with less precision (the field of observation of an object
ogy-related error, whereby it is not possible to transfer re- on theroad is greater in theelderly than in theyoung).
sults to real situations. However, conscious orientation of Theauthors believe that theresults justify continuation of
attention seems to affect object conspicuity. Thesearched that research.
object is 3 times as visible as theobject accidentally pres- Fiedler and Glckner have measured thenumber of fixa-
ent in thevisual field[30]. tions on avisual target, and percentage of thenumber of
total fixations throughout the measurement period [41].
RELEVANT EYE-TRACKINGINDICATORS Using thenon-contact FaceLab apparatus, they have also
FORUSE IN THE STUDY measured the average time of fixation and distribution
OFVISUALSTRATEGIESONDRIVING of long, medium and short fixations. Decomposition of
Most common indicators in theassessment of visual strate- thefield of view into regions of interests was conducted
gies include: thenumber of fixations, average fixation time, depending on thecapacity of theeye-tracking system.
and exploration areas of thefunctional vision field. These In thestudies of thedynamics of risk-taking in acomputer
parameters were measured as early as in the1970s[10], game they have also measured thechanges in thewidth
although theaccuracy attainable at that time was not sat- of the pupil by the PERCLOS (Percent Eyelid Closure)
isfactory. They are still in use considering that they can be system and the time of decision-making in terms of dif-
optimized, primarily by statistical means. ference between theappearance of astimulus and theas-
Muttart et al. [39] employing a Mobile-Eye miniatur- sociated response. Theauthors have recorded eye move-
ized contact system used those particular parameters in ments at120Hz sampling rate frequency and aprecision
comparative studies of visual strategies of drivers of cars of thelocation of theeye movement of about0.45. Fixa-
and motorcycles during a left turn or U-turn manoeu- tion was identified with aprecision of30px, and therec-
vre. They have found that motorcyclists were developing ognition threshold was set at50ms. Thereaction time was
amuch richer visual strategy than car drivers, as reflected measured with accuracy up to5ms. Theresults confirm
by agreater number of fixations, especially on theobjects that therate of change of these indicators can be useful

IJOMEH 2015;28(6) 949


REVIEW PAPER B. KAPITANIAK ET AL.

for theevaluation of mental workload[42] although theal- on theroad in order to maintain thecorrect position on
gorithms based on a pupil diameter are not sufficiently the lane during driving around a curve [46]. The more
robust if thelight conditions change. distant of those2sources is used as atarget point to pre-
Lang and Qi have proposed an algorithm for detec- dict thecurvature of theroad. This source on theroad
tion of drivers fatigue, based on measurements of is in thedrivers field of view at thedistance that theve-
changes in apupil diameter(PERCLOS), and therate of hicle travels during theoptimal0.751s and is used for
blinking[42]. anticipation, allowing the driver to predict curvature
Assessment of theproportion of eye movements to head of the road. The other, less distant source, preceding
movements during a precise location of a visual target the driver by about 0.5 s is used as a re-control (feed-
was conducted as early as in the 1970s. Zangemeister back) mechanism that allows the driver to fine-tune
and Stark observed an offset (or desynchronisation) themanoeuvre. Thestudies have shown that this mecha-
of 2 movements [43]. According to Doshi and Trivedi, nism is present in most people, regardless of experience
a separate analysis of eye movement and head move- in driving, but its accuracy is better in experienced driv-
ment allows for a more accurate assessment of visual ers. This mechanism occurs in most management tasks,
strategies[44]. such as driving abicycle, horse riding or simulated driv-
It is generally recognized that errors of visual attention are ing on acomputer[47,48].
responsible for alarge proportion of traffic accidents[24]. However, Kandiletal. [49,50] point out to adiscrepancy
Recording and analyzing eye movements provide impor- between theresults of research conducted on simulators
tant elements for understanding thenature of thetask of and real-time driving. Theresults compared with thelit-
vehicle driving. These results make it possible to optimize erature suggests that thestrategy of thereal driving is most
driver training strategy and prevent accidents. Thecited au- often a target point (described above and in the Dis-
thors suggest that in thesimplest case of driving, adrivers harts study), while during simulated driving, atarget sam-
visual fixation on straight stretches of theroad can be de- pling strategy is more frequently observed, which involves
scribed as focusing on theexpansion point (target),i.e.,on progressively moving virtual point according to themove-
that section of thevisual field in front of adriver where ob- ment of thevehicle[49,50].
jects appear to be stationary, with rare gazes on thetraffic The main problem in conducting research on visual strat-
signs and markers on theedge of theroad. It is clear that egy is to identify and monitor the requirements of driv-
a dangerous event usually causes a significant increase ers as independent variables[51]. Potential confounding
in theduration of fixation. factors include theincreased visual requirements, such as
Kito et al. have found that the amplitude of eye move- theincrease in visual complexity, and increase in cognitive
ments is significantly broader when driving large vehicles demands, such as increased processing requirements of
(truck, bus) than during driving asmall car[45]. Fixation aparticular stimulus because of its importance in thesitu-
positions are observed at approximately 5060 relative ation. It seems that the overall increase in the require-
to thecentral axis of adrivers sight(i.e.,5060, both to ments and complexity of thetask tends to reduce theav-
left and right of theaxis). Some of them are targeted at erage duration of fixation and increases thenumber and
zone020 relative to that axis. frequency of fixations.
Dishart and Land suggest that experienced drivers re- According to Duchowski, older people are less accurate in
ceive visual information from 2 sources of information capturing scenes viewed while driving[52]. During night

950 IJOMEH 2015;28(6)


EYE-TRACKING AND TESTING OF DRIVERS REVIEW PAPER

driving, the difference is even greater. During daytime condition (9574.5 ms vs. 6523.4 ms). The people talk-
driving, visual distractions, such as billboards located on ing on a mobile had fewer fixations on the hazardous
theroadside of ahighway, increase thefrequency of er- vehicles while the hazardous event occurred (118 ms
rors, which is not thecase at night. There is astrong cor- vs.186 ms). The total time spent fixating on the hazard-
relation between thenumber of fixations and theresponse ous vehicles during the event was longer for the people
time. Older people need more fixation than younger ones in the driving-only condition (mean (M) = 4755.3 ms)
to follow thesame scene, as well as more fixation is neces- than for those talking on a mobile (M = 3514.8 ms).
sary to recognize thescene in the circuimstances of poor Many authors have applied theanalysis of eye movements
visibility (fog, rain, etc.). Similar results have been ob- to assess thelevel of drivers fatigue. Merat and Jamson
tained by Fieldesetal., who have found that older people have found differences in thelevel of fatigue by analyzing
respond more slowly than younger people, but also their thepupil diameter using PERCLOS[57]. Schleicheretal.
visual strategy is qualitatively different and can be judged determines thelevel of fatigue based on thefrequency of
as less effective[40]. blinking and saccade characteristics[58].
The results of experiments in which gaze is monitored
in simulated driving conditions indicate that the prop- CONCLUSIONS
er task visibility depends mainly on the habits acquired The present review of literature indicates that themeth-
by the driver during driving. The findings suggest that od of tracking eye movements is currently themethod of
thestandards of fixation and visual control in individuals choice for thestudy of cognitive strategies, in particular vi-
with normal visual skills are acquired by learning[53]. sual strategy. Currently available techniques allow theuse
The studies investigating effects of mental activity of non-contact methods in theconditions of field test, both
while driving suggest the desirability of an awareness during simulated driving and in real traffic conditions.
campaign about the risks of concentrating on their Theresults enable aquantitative assessment of objective
own thoughts unrelated to theprocess of driving[54]. parameters of thestrategy and comparing these strategies
Numerous studies on the effects of performing dual- under different experimental conditions.
task procedure, involving driving a vehicle and talk- Although the number of publications on the analysis of
ing on a mobile phone, on the efficiency of perform- eye movements while driving increases, it appears that
ing themain (driving) task indicate that this efficiency much of theresearch in this area is not published due to
drops very seriously, which is reflected in theincreased their direct use in concept vehicles.
number of violent braking episodes, lack of control, It is advisable that theprospective development of that re-
neglecting rear mirror view and impaired strategies for search includes, in particular, basic research intended to
recognizing objects on theroad[55]. explain learning mechanisms that are necessary for under-
The study of Balk et al. [56] realised on a low fidelity standing the functioning of thehuman cognitive system.
driving simulator has shown that the percentage of This perspective is particularly attractive for thefuture of
fixations on the vehicles involved in hazardous events ergonomics, as it would provide aneurophysiological ba-
throughout the test was greater for the driving-on- sis for a better understanding of the cognitive processes
ly condition than driving and talking on a mobile necessary for theproper design of theinterface for infor-
(40% vs. 29%). The mean time spent looking at the mation exchange between aman and themachine oper-
hazardous vehicles was greater for the driving-only atedby him.

IJOMEH 2015;28(6) 951


REVIEW PAPER B. KAPITANIAK ET AL.

REFERENCES Q J Exp Psychol. 2009;62(8):1457506, http://dx.doi.org/


1. Wade NJ. Pioneers of eye movement research. i-Percep- 10.1080/17470210902816461.
tion.2010;1:3368, http://dx.doi.org/10.1068/i0389. 15. Shin D, Sakai H, UchiyamaY. Slow eye movement detection
2. Gilliland AR. Photographic methods for studying reading. can prevent sleep-related accidents effectively in asimula
Vis Educ.1921;2:216. ted driving task. J Sleep Res. 2011;20(3):41624, http://dx.
3. Hartridge H, Thompson LC. Methods of investigating eye doi.org/10.1111/j.1365-2869.2010.00891.x.
movements. Br J Ophthalm. 1948;32:58191, http://dx.doi. 16. Levy-Schoen A. [Eye movements exploration]. Lanne Psy-
org/10.1136/bjo.32.9.581. chologique. 1967;67(2):56999, http://dx.doi.org/10.3406/
4. Shackel B. Note on mobile eye viewpoint recording. psy.1967.27583. French.
J Opt Soc Am. 1960;59:7638, http://dx.doi.org/10.1364/ 17. Miura T. Coping with situational demands: Astudy of eye
JOSA.50.000763. movements and peripheral vision performance. In:GaleAG,
5. Mackworth NH, Thomas EL. Head-mounted eye-mark- FreemanMH, HaslegraveCM, SmithP, TaylorSP, editors.
er camera. J Opt Soc Am. 1962;52:7136, http://dx.doi. Vision in vehicles. Amsterdam: Elsevier, North Holland
org/10.1364/JOSA.52.000713. Press;1986.p.20516.
6. Young LR, Sheena D. Survey of eye movement recording 18. Salvucci DD, Goldberg JH. Identifying fixations and sac-
methods. Behav Res Methods Instrum.1975;7(5):397429, cades in eye-tracking protocols. Proceedings of the Eye
http://dx.doi.org/10.3758/BF03201553. Tracking Research and Applications Symposium. New
7. Soliday M. Drivers eye movements: Aliterature review. Ap- York:ACMPress;2000.p.718.
pendixD. In:JohnsTR, AllenJA. Driver license and testing. 19. Horrey WJ, Wickens CD. In-vehicle glance duration dis-
Highway Safety Research Center, University of North Caro- tributions, tails, and model of crash risk. Transportation
lina, North Carolina;1971.p.1523. research record, 2018. Washington (DC): Transportation
8. Cohen AS, Studach H. Eye movements while driving cars Research Board of theNational Academies;2007.p.228.
around curves. Percept Mot Skills. 1977;44:6839, http:// 20. Martens MH, Fox MRJ. Do familiarity and expecta-
dx.doi.org/10.2466/pms.1977.44.3.683. tions change perception? Transp Res Part F Traffic Psy-
9. Shinar D, Mcdowell ED, RockwellTH. Eye movements in chol Behav. 2007;10:47692, http://dx.doi.org/10.1016/j.trf.
curve negotiation. Hum Factors.1977;19:6371. 2007.05.003.
10. Neboit M. [Simulation and learning driving]. LeTravail Hu- 21. Kaczmarek R. [Eye command to aid communication and
main.1978;41(2):23949. French. environmental control for the disabled people]. Motricit
11. Mourant RR, Rockwell TH. Strategies of visual search by nov- Crbrale.1992;13:2430. French.
ice and experienced drivers. Hum Factors.1972;14(4):32535. 22. Galley N. Theevaluation of theelectrooculogram as apsy-
12. Rayner K. [Eye movements in reading and information pro- chophysiological measuring instrument in the driver study
cessing:20years of research]. Psychol Bull.1998;124(3):372 of driver behaviour. Ergonomics.1993;36(9):106370, http://
422, http://dx.doi.org/10.1037/0033-2909.124.3.372. French. dx.doi.org/10.1080/00140139308967978.
13. Ross J, Burr DC, MorroneMC. Suppression of themagno- 23. Richardson DC, Michael J, SpiveyMJ. Eye-tracking: Char-
cellular pathways during saccades. Behav Brain Res.1996; acteristics and methods, research areas and applications.
80:18, http://dx.doi.org/10.1016/0166-4328(96)00012-5. In: Wnek G, Bowlin G, editors. Encyclopedia of biomate-
14. Rayner K. Eye movements in reading and atten- rials and biomedical engineering. London: Taylor & Fran-
tion in reading, scene perception, and visual search. cis;2004. p.170.

952 IJOMEH 2015;28(6)


EYE-TRACKING AND TESTING OF DRIVERS REVIEW PAPER

24. Chapman PR, Underwood G. Visual search of dynamic basic and clinical aspects of vertigo and dizziness: Ann
scenes: Event types and the role of experience in view- NYAcad Sci.2009;1164:18893, http://dx.doi.org/10.1111/
ing driving situations. In:UnderwoodG, editor. Eye guid- j.1749-6632.2008.03714.x.
ance in reading and scene perception. Oxford: Elsevi- 34. Pottier A, Neboit M. [The ergonomic work analysis
er;1989.p.36994. by the study of visual exploration]. Toulouse Ed Oct-
25. Drewes H. Eye gaze tracking for human computer interac- ares;1995.p.1162. French.
tion [dissertation]. Mnchen: LFE Medien-Informatik der 35. Trivedi MM, Gandhi T, McCallJ. Looking-in and looking-
Ludwig-Maximilians-Universitt;2010. out of a vehicle: Computer-vision-based enhanced vehicle
26. Ould AM, Perreira Da Silva M, Courboulay V. [A history safety. IEEE Intell Transp Syst. 2007;8(1):10820, http://
of eye gaze tracking.CCSD CNRS, 2008; HAL-00215967] dx.doi.org/10.1109/TITS.2006.889442.
[cited 2014 Apr 30]. Available from: http://ccsd.cnrs.fr. 36. Van Loon EM, Khashawi F, Underwood G. Visual strate-
French. gies used for time-to-arrival judgments in driving. Percep-
27. Layerle JF, Savatier X, Ertaud JY, Mouaddib EM. [De- tion.2010;39(9):121629, http://dx.doi.org/10.1068/p6067.
sign of acatadioptric sensor for an optimal observation of 37. Viviani P. Eye movements in visual search: Cognitive, per-
thedriver and theroad scene]. ORASIS09 Congrs des ceptual and motor aspects. In:KowlerEE, editor. Eye move-
jeunes chercheurs en vision par ordinateur;2009 Jun812; ments and their role in visual and cognitive processes. Am-
Trgastel, France. HAL/INRIA;2009. French. sterdam: Elsevier Science;1990.p.35393.
28. Sanders AF. Processing information in thefunctional visual 38. Edquist J, Horberry T, Hosking S, Johnston I. Effects of
field. In:dYdewalleG, vanRensbergenJ, editors. Percep- advertising billboards during simulated driving. Appl Er-
tion and cognition: Advances in eye movement research. gon. 2011;42(4):61926, http://dx.doi.org/10.1016/j.apergo.
Amsterdam: North Holland;1993.p.322. 2010.08.013.
29. Boersema T, Harm JG, ZwagaHJG. Theinfluence of adver- 39. Muttart JW, Peck LR, Guderian S, Bartlett W, Ton LP,
tisements on the conspicuity of routing information. Appl KaudererC,etal. Glancing and stopping behavior of mo-
Ergon. 1985;16(4):26773, http://dx.doi.org/10.1016/0003- torcyclists and car drivers at intersections. Transp Res
6870(85)90090-0. Rec.2011;2265:818, http://dx.doi.org/10.3141/2265-09.
30. Brmond R. The conspicuity of traffic signs in complex 40. Fildes B, Charlton J, MuirC, KoppelS. Driving responses
backgrounds. RapportCIE.2003;137. of older and younger drivers in a driving simulator. Annu
31. Itti L, Koch C. A saliency-based search mechanism Proc Assoc Adv Automot Med.2007;51:55972.
for overt and covert shifts of visual attention. Vision 41. Fiedler S, Glckner A. The dynamics of decision mak-
Res. 2000;40:1489506, http://dx.doi.org/10.1016/S0042- ing in risky choice: An eye-tracking analysis. Front Psy-
6989(99)00163-7. chol.2012;3:335, http://dx.doi.org/10.3389/fpsyg.2012.00335.
32. Cole BL, Jenkins SE. The effect of variability of back- 42. Lang L, Qi H. The study of driver fatigue monitor algo-
ground elements on the conspicuity of objects. Vision rithm combined PERCLOS and AECS. International
Res. 1984;24:26170, http://dx.doi.org/10.1016/0042-6989 Conference on Computer Science and Software Engineer-
(84)90129-9. ing; 2008 Dec 1214; Washington (DC), USA. Los Alami-
33. Einhuser W, Schumann F, Vockeroth J, Bartl K, Cerf M, tos (CA), Washington, Tokyo: Conference Publishing Ser-
Harel J, et al. Distinct roles for eye and head movements vice, IEEE Computer Society; 2008, http://dx.doi.org/10.
in selecting salient image parts during natural exploration, 1109/csse.2008.771.

IJOMEH 2015;28(6) 953


REVIEW PAPER B. KAPITANIAK ET AL.

43. Zangemeister WH, Stark L. Gaze latency: Variable interac- perception. Oxford: Elsevier;1998.p.395418, http://dx.doi.
tions of head and eye latency. Exp Neurol.1982;75(2):389 org/10.1016/B978-008043361-5/50019-5.
406, http://dx.doi.org/10.1016/0014-4886(82)90169-8. 52. Duchowski AT. Abreadth-first survey of eye tracking app
44. Doshi A, Trivedi MM. Head and eye gaze dynamics dur- lications. Behav Res Methods Instrum Comput. 2002;34
ing visual attention shifts in complex environments. J Vis. (4):45570, http://dx.doi.org/10.3758/BF03195475.
2012;12(2):9. 53. Land MF, Lee DN. Where we look when we steer. Na-
45. Kito T, Haraguchi M, Funatso H, Sato T, Kondo M. Mea- ture.1994;369:7424, http://dx.doi.org/10.1038/369742a0.
surements of gaze movements while driving. Percept 54. Duchowski A, Medlin E, CourniaN, MurphyH, Gramopad-
MotSkills. 1989;68(1):1925, http://dx.doi.org/10.2466/pms. hyeA, NairS,etal.3Deye movement analysis. Behav Res
1989.68.1.19. Methods Instrum Comput. 2002;34(4):118, http://dx.doi.
46. Dishart DC, Land MF. The development of the eye movement org/10.3758/BF03195486.
strategies of learner drivers. In: Underwood G, editor. Eye guid- 55. Muttart JW, Fisher DL, Knodler M, Pollatsek A. Driving
ance in reading and scene perception. Oxford: Elsevier; 1998. simulator evaluation of driver performance during hands-
p. 41930, http://dx.doi.org/10.1016/B978-008043361-5/50020-1. free cell phone operation in awork zone: Driving without
47. Land MF. Predictable eye-head coordination during driv aclue. Washington: Transportation Research Board Annual
ing.Nature.1992;359(6393):31820. Meetings; 2007.p.914.
48. Lehtonen E, Lappi O, Kotkanen H, Summala H. Look- 56. Merat N, Jamson AH. Theeffect of three low-cost engineer-
ahead fixations in curve driving. Ergonomics. 2013;56(1): ing treatments on driver fatigue: Adriving simulator study.
3444, http://dx.doi.org/10.1080/00140139.2012.739205. Accid Anal Prev. 2013;50:815, http://dx.doi.org/10.1016/
49. Kandil FI, Rotter A, LappeM. Driving is smoother and more j.aap.2012.09.017.
stable when using thetangent point. JVis.2009;9(1):111. 57. Balk S, Moore K, Steele J, Spearman W, Duchowski A.
50. Kandil FI, Rotter A, Lappe M. Car drivers attend to dif- Mobile phone use in a driving simulation task: Differenc-
ferent gaze targets when negotiating closed vs.open bends. es in eye movements. J Vis. 2006;6(6):872, http://dx.doi.
JVis.2010;10(4):24.111, http://dx.doi.org/10.1167/10.4.24. org/10.1167/6.6.872.
51. Crundall DE, Underwood G, Chapman PR. How much 58. Schleicher R, Galley N, BriestS, GalleyL. Blinks and sac-
do novice drivers see? The effects of demand on visual cades as indicators of fatigue in sleepiness warnings: Look-
search strategies in novice and experienced drivers. In:Un- ing tired? Ergonomics. 2008;51(7):9821010, http://dx.doi.
derwood G, editor. Eye guidance in reading and scene org/10.1080/00140130701817062.

This work is available in Open Access model and licensed under a Creative Commons Attribution-NonCommercial 3.0 Poland License http://creativecommons.org/
licenses/by-nc/3.0/pl/deed.en.

954 IJOMEH 2015;28(6)

You might also like