Conference PaperPDF Available

Using Non-Calibrated Eye Movement Data To Enhance Human Computer Interfaces

Authors:

Abstract and Figures

Eye movement may be regarded as a new promising modality for human computer interfaces. With the growing popularity of cheap and easy to use eye trackers, gaze data may become a popular way to enter information and to control computer interfaces. However, properly working gaze contingent interface requires intelligent methods for processing data obtained from an eye tracker. They should reflect users' intentions regardless of a quality of the signal obtained from an eye tracker. The paper presents the results of an experiment during which algorithms processing eye movement data while 4-digits PIN was entered with eyes were checked for both calibrated and non-calibrated users.
Content may be subject to copyright.
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
adfa, p. 1, 2011.
© Springer-Verlag Berlin Heidelberg 2011
Using Non-Calibrated Eye Movement Data To Enhance
Human Computer Interfaces
Pawel Kasprowski, Katarzyna Harezlak
Institute of Informatics
Silesian University of Technology
Gliwice, Poland
pawel.kasprowski@polsl.pl, katarzyna.harezlak@polsl.pl
Abstract. Eye movement may be regarded as a new promising modality for
human computer interfaces. With the growing popularity of cheap and easy to
use eye trackers, gaze data may become a popular way to enter information and
to control computer interfaces. However, properly working gaze contingent in-
terface requires intelligent methods for processing data obtained from an eye
tracker. They should reflect users intentions regardless of a quality of the signal
obtained from an eye tracker. The paper presents the results of an experiment
during which algorithms processing eye movement data while 4-digits PIN was
entered with eyes were checked for both calibrated and non-calibrated users.
1 Introduction
The usefulness of the eye movement analysis was confirmed in research in many
areas of interests. It may be used for example in advertisements developing, sociolo-
gy, medicine and cognitive studies [3][6]. Recently, a lot of attention has been fo-
cused on possibilities to use eye movement for enhancing human-computer interfaces.
Using gaze information as a new input device in a way similar to mouse seems to be
the promising technique, making cooperation with computers even easier for unexpe-
rienced users. Nevertheless, the eye movement processing still faces a lot of usability
problems so a lot of effort must be done to make this technique really user friendly.
One of the examples is a so called Midas touch problem [7], which addresses the
difficulty to decide when user looks at a button if he wants to click it or just to read its
caption.
One of the most important obstacles in making eye movement based interfaces ro-
bust and convenient is the necessity to calibrate an eye tracker for each user before
any usage [11]. The aim of the studies presented in this paper was to check whether it
is possible to utilize information obtained from an eye tracker without prior calibra-
tion done by the user that is being measured. One of the simplest tasks - the PIN en-
tering - was taken into consideration and these studies are based on the research dis-
cussed in [9].
The main contribution of the research presented in the paper is the introduction of
the idea to shorten eye tracking sessions by carrying out the same calibration for vari-
ous users. Thus, the novel so called regression based algorithm was implemented and
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
compared to an intuitive distance based algorithm. The correctness of the analysed
task realization for both cases and factors influencing the results were analysed.
All classic experiments based on eye tracking methods are conducted in accord-
ance with a commonly used schema. Each trial starts with a calibration process. The
aim of that step is to find the correlation between coordinates of user’s gaze point and
coordinates represented in an eye tracker system. During a calibration users are asked
to move their eyes over a screen as a reaction to a presented stimulus. Dependent on
an experiment type there may be various stimuli used, yet the most popular is a point
jumping over a screen. Each change of the point position triggers eyes movement.
Recorded eye movement samples correspond to a given point on a screen and a user is
expected to keep the focus in the same point for a while, long enough to collect such
set of samples, which ensure good adjustment of a screen and an eye tracking system
coordinates. The time of a single point presentation is usually set within scope of 2
3 seconds [4]. A number of point’s locations and their dispersion on the screen are
other issues [8][1]. It is obvious that the higher quality may be achieved for more
points, yet taking user’s convenience into consideration the lowest possible number is
better solution. The reason is time required to perform an eye tracker calibration too
long can be wearisome for participants, discouraging them for the involvement in core
experiments.
The main problem of the calibration is that it must be repeated before every trial as
it depends on an environment used in experiments and on characteristic features of a
particular user. As the calibration process is rather cumbersome for a user, the idea of
the paper was to check if, for some human-system interactions, which do not require
highly accurate eye tracker’s adjustment, it is possible to omit calibration step and
still achieve satisfactory results. Tasks regarding entering a PIN, in which focusing
eyes on a specific area is sufficient to determine a digit, can be taken as an example.
Such task may be used to lock and unlock computer screen with eyes or to enter PIN
at ATM [10]. According to [2] it prevents shoulder surfing attack and is generally
more difficult to forge.
2 The experiment
The eye movements were registered using the Eye Tribe - an eye tracking system
working with sampling rates 60 Hz. The accuracy and spatial resolution declared by
manufacturer equals 0.5° 1° and 0.1° respectively. The eye tracker was placed be-
low the screen. A PC computer was used to control the experiment (show stimulus,
control the eye tracker and save recordings). The users were sitting centrally at a dis-
tance of 60 cm.
The experiment consisted of two parts. During the first part one participant was
calibrated using a classic scenario with 9 points evenly distributed on the screen. It
lasted for about 20 seconds. Then a screen with evenly distributed circles with digits
from 0 to 9 was displayed and the participant was asked to look at a digit and press a
trigger button, then look at subsequent digit and press the trigger and so on. The PIN
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
was a four digits sequence for which every two subsequent digits were always differ-
ent. After four trigger clicks the attempt (a trial) was saved.
The second (and more interesting) part of the experiment started when some num-
ber of different participants were asked to enter their PINs with eyes in the same
manner, but this time without any calibration. These participants’ eye movements
were registered using a calibration function built for the first user. Eye movement data
for every PIN entering was saved (later referred to as a trial).
There were 802 trials collected including 204 with own calibration (called ‘cali-
brated’) and 598 without own calibration (called ‘non-calibrated’). 41 participants
took part in the experiment.
To examine samples gathered during the experiments two own developed methods
were used. Both of them are based on sets of fixations extracted from eye movement
signal and are described in details in the subsequent section.
3 Method
The purpose of the PIN extraction algorithm described in this section was to obtain
information about a sequence of digits pointed with eyes from the recorded eye
movement data. In the algorithm two phases may be distinguished :
Extraction of fixations,
Assigning fixations to digits on the screen.
3.1 Extraction of fixations
Typical eye movement signal consists of two events: fixations - when eye is rela-
tively still and the brain acquires information from the scene; and saccades - a rapid
movement when an eye position changes to another fixation. The extraction of fixa-
tions from a raw eye movement signal may be done using different algorithms
[12][13]. It was our own implementation of one of the most popular dispersion-
threshold algorithm (I-DT) used in this work.
At first the algorithm classifies each eye movement sample according to a simple
rule: if the distance among this sample and five previous samples is less than a speci-
fied threshold (Th) the sample is classified as a potential part of a fixation (F) and it is
classified as a potential part of a saccade (S) otherwise. In the next step all neighbor-
ing F-points are gathered together as potential fixations. Every fixation has four at-
tributes: its start time, x and y coordinates of its center and the fixation duration. The
subsequent steps convert this preliminary list of fixations into the final list using dif-
ferent techniques for fixation merging and removing. All details of the algorithm are
presented in [5].
The value of threshold parameter (Th) started from 0.2 deg. and was increased by
0.2 deg. until one of following conditions was met: I-DT algorithm returned exactly
four fixations or threshold value reached 8 deg.
In the latter case the trial was rejected and no further analyses were done.
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
3.2 Discovering chosen digits
After extracting the four most dominant fixations, it was assumed that these fixa-
tions occurred while a person was looking at specific digits. The next task was to
discover which digits were pointed with eyes. There were two different methods used
in the research to pair fixations with proper digits.
Distance based approach.
The first and the most obvious - method divides a screen into regions of interest
(ROIs) using digits locations as points in Voronoi diagram. As a result every fixation
is classified as a digit which is the closest one for this fixation. It may happen that a
fixation location is almost in the middle between two digits (near a boundary of a
Voronoi cell). As a result one of the digits must always be chosen but we may expect
that the choice is somehow random in such case. Therefore, an additional step and an
additional parameter: proximity coefficient (PCF) were introduced. After finding the
closest digit for a fixation, it is checked whether a distance between that digit and the
fixation multiplied by PCF is still lower than distances between the fixation and all
other digits:
        (1)
where F is fixation’s location and Di is a location of digit i.
If this condition does not hold for any fixation in a sequence, the whole trial is re-
jected. Obviously, for PCF equal to one there are no rejections (there is always one
minimal distance) and as its value increases the number of rejections increases as
well. For instance PCF equal to 2 means that the distance between the closest digit
and the fixation must be twice lower than the distances between that fixation and all
other digits.
To check whether a simple, one point calibration is enough to improve results, the
additional assumption was added that the first digit of PIN is known. Therefore, the
extended version of the algorithm introduces the additional step: before any classifica-
tion all fixations are shifted in space so that the first fixation is positioned exactly in a
location of the first digit. The confirmation of usefulness of such activity may be sub-
sequently used for one point calibration displayed for example in the middle of the
screen.
Regression based approach.
The next method was using slightly different approach. The basic assumption was
that the regression model for a correctly adjusted PIN should provide the lowest error.
Thus, the algorithm starts with building regression models that map 4 fixations into 4
digits for every possible combination of PIN digits. There are 7290 possible combina-
tions when assumption that subsequent digits are always different holds. For every
such model new fixations’ positions are calculated and Mean Square Error between
these new and correct positions of PIN’s digits are calculated. At the end there is a list
of possible PIN numbers with MSE for every PIN available. The PIN with the lowest
error is chosen as a correct one.
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
The only problem in the algorithm described above is which regression function
should be chosen to obtain reliable results. Usually, second degree polynomial func-
tion is used for eye trackers calibration [1] but such model is too precise. For 4 points
it is able to build a function that maps given fixations to any sequence of digits with
almost no errors. Therefore, it was a first degree polynomial function used to evaluate
new values for X and Y independently:
      (2)
      (3)
The Levenberg Marquardt algorithm was used to calculate coefficients for each 4
fixations 4 digits pair. Because we were not interested in mirror mapping of PIN
numbers, an additional assumption that coefficients Ax and Ay must be positive num-
bers was made.
Similarly to the proximity coefficient in the distance based method, it was neces-
sary to add possibility to allow for rejection of trials for which values found are not
reliable. Therefore, an additional min_error (MER) parameter was introduced. If the
lowest value of MSE for the trial is higher than MER, the trial is rejected as unrelia-
ble.
To make a fair comparison to the distance based approach that uses information
about the first digit, there was also a version of the algorithm evaluated that calculates
models only for PINs starting with a known digit.
4 Results
The most obvious results that may be taken into consideration is the absolute accu-
racy (ABS), which is measured as a ratio between the number of trials with PIN found
correctly and the number of all trials. However, there are two more detailed factors
that may be used when analyzing results of the algorithms described in the previous
section. At first, each algorithm rejects some number of trials for which it assumes
that recognition is impossible. So, the first factor to be analyzed is an acceptance rate
(ACR). This factor is influenced by PCF and MER parameters accordingly to the
algorithm used. Then, for all remaining trials, PIN is evaluated. The number of PINs
found correctly to the number of all evaluated trials is defined as a correctness rate
(CRR).
Ideally, both ACR and CRR should be 100%. However, it can be expected that
both factors are dependent on each other when the acceptance rate decreases, the
remaining samples are of better quality and the correctness increases. And when the
acceptance rate increases, more low quality samples are taken into account during the
next step, which may result in lower correctness rate. Tuning of this two factors de-
pends on the purpose of the trial (see Conclusion for examples). Obviously, the abso-
lute accuracy (ABS) is the result of multiplication of the two above factors (ABS =
ACR*CRR).
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
As it was presented in the previous section, the first rejection takes place after the
‘extraction of fixations’ step. All trials, for which it was impossible to find exactly
four fixations are rejected. The next step when trials may be rejected depends on the
algorithm used. For the distance based algorithm the acceptance rate depends on the
proximity coefficient (PCF). As it was described in the previous section, increasing
PCF decreases the number of accepted samples. For the regression based algorithm
the min_error coefficient (MER) may be tuned to reject dubious trials. If MER is
high, all samples are accepted and as it decreases, the acceptance rate (ACR) decreas-
es as well.
To illustrate described dependency, the ACR and CRR values for trials when the
regression based algorithm was used with different values of min_error (MER) pa-
rameter was presented in Fig 1.
Fig. 1. ACR and CRR values for different min_error (MER) in regression based algorithm
The results obtained for both algorithms and both types of samples are shown in
following tables. Table 1 presents values of the acceptance rate (ACR) and the cor-
rectness rate (CRR) for calibrated and non-calibrated trials, for the distance based
(DIST) algorithm with different values of the proximity coefficient (PCF).
Table 1. CRR and ACR for different PCF for distance based algorithm
Calibrated
Non-calibrated
PCF
CRR
ACR
CRR
ACR
1
95%
94%
58%
68%
1.1
95%
94%
66%
57%
1.2
95%
93%
75%
46%
As it can be seen, the results for calibrated trials are quite good and stable for dif-
ferent values of PCF and the results are significantly better than for non-calibrated
trials. As it could be expected, higher value of PCF increases the correctness (CRR)
but in the same time decreases the acceptance (ACR). For PCF>=1.2 more than a half
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
of non-calibrated trials is rejected. The best value of ABS for non-calibrated trials is
only 39% while it is about 90% for all PCF values, when only calibrated trials are
taken into consideration.
The results for the regression based algorithm (REGR) are presented in Table 2.
They were calculated for different values of min_error (MER) parameter.
Table 2. CRR and ACR for different MER values for regression based algorithm
Non-calibrated
MER
CRR
ACR
CRR
ACR
1
74%
96%
56%
94%
0.5
86%
94%
66%
91%
0.1
94%
86%
74%
80%
0.08
96%
83%
81%
73%
0.06
96%
82%
85%
61%
0.04
97%
75%
90%
44%
It is visible that the results for calibrated trials are worse than for DIST algorithm
with ABS about 80%. However, the results for non-calibrated trials for the regression
based algorithm are significantly better than for the distance based one, with ABS
reaching 60% for MER=0.5. The algorithm is especially efficient in rejecting low
quality trials. For instance, 74% correctness (CRR) was achieved for the acceptance
rate (ACR) 80%, while for DIST algorithm the same correctness rate was achieved
for ACR amounting only to 46%.
The next research question was how the simplest possible, one point calibration
can improve the results. Because there were only trials with four points available the
only way to check it was to assume that the first digit of PIN is known. For DIST
algorithm it resulted in shifting fixations so that the first fixation overlapped the first
(known) digit (see Method section for details). For REGR algorithm only PIN num-
bers starting with the known digit were considered as candidates (see Method section
as well).
Table 3. Results achieved for DIST algorithm with the first fixation shift.
Calibrated
Non-calibrated
PC
CRR
ACR
CRR
ACR
1
96%
95%
76%
83%
1.1
96%
94%
81%
77%
1.2
96%
94%
85%
70%
1.3
98%
93%
89%
62%
1.4
98%
91%
91%
56%
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
When considering DIST algorithm (Table 3) the results for the calibrated trials are
better for the case of fixation shifting but the difference is not significant (ABS is
equal about 91% in most cases). However, the results for non-calibrated samples are
significantly better with 63% for ABS, in the best case comparing to 39% for tests
without shifting.
Table 4. Results for REGR algorithm with the first digit known
Calibrated
Non-calibrated
MER
CRR
ACR
CRR
ACR
0.5
87%
94%
69%
90%
0.1
95%
85%
83%
74%
0.08
96%
83%
88%
69%
0.06
98%
80%
91%
58%
0.04
98%
74%
92%
43%
The results for the regression based algorithm (Table 4) did not improve outcomes
significantly when only PINs with a correct first digit were taken into account. Such
condition reduced the number of PINs for which models were calculated ten times
(from 7290 to 729) but it did not affect algorithms performance.
Fig. 2. ACR and CRR for different algorithms and calibrated (cal) and non-calibrated (nc) trials
The comparison of the algorithms and the sets was presented in Fig 2. The distance
based algorithm performed very well for calibrated data (dist_cal and dist_cal_shift)
while its results were very unsatisfactory for non-calibrated data (dist_nc). The re-
gression algorithm was not as good as the distance based one for calibrated data
(regr_cal) but it outperformed it for non-calibrated one (regr_nc). Adding infor-
mation about the first digit improved outcomes of the distance based algorithm
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
(dist_nc_shift) but even with this information it is not better than the normal regres-
sion based outcome (regr_nc). As it was shown in Table 4, adding information about
the first digit of PIN did not improve significantly the results for the regression based
algorithm so it was not included in Fig 2.
5 Conclusions
The findings of the research can be divided into two groups. First of them regards
trials proceeded by the per-user calibration. The results obtained for such recordings
confirmed the possibility of using eyes for providing information of PIN type. Anoth-
er conclusions may be drawn from outcomes obtained for various scopes of the intro-
duced parameters - proximity coefficient (PCF) and min_error (MER). They show to
what extent the size of the area of interests can be reduced not to decrease the effi-
ciency of the method.
The second group of the findings concerns the problem of omitting a calibration
process. The experiments presented in this paper showed that it is possible to use eye
tracker as a pointer for simple and well defined tasks even without a prior per-user
calibration. It is possible if such task does not require point to point gaze mapping, yet
point to area of interests adjustment is acceptable. The studies of using eyes for PIN
providing fulfil this requirement. The results obtained for the non-calibrated trials are
worse than for the calibrated ones, however values of the analyzed factors indicated
that in most cases proper values could be obtained.
A novel regression based algorithm was introduced and it was shown that it outper-
forms the distance based one for the non-calibrated samples. Additionally, it may be
tuned for various types of interfaces using a min_error parameter. When the correct-
ness of recognition is important, the min_error value may be increased and it was
shown that results become more reliable (in sake of higher rejection rate). Such sce-
nario may be useful for instance for gaze pointing of PIN at ATM when we want to be
sure that PIN entered is correct even if the user is forced to enter it several times due
to rejections. On the other hand there are applications in which an approximated gaze
position is enough and rejections are rather undesirable for an interface to be fluent.
That is the case of for instance interactive games. For such applications min_error
value may be low, resulting in lower rejection but also with lower overall accuracy.
Providing opportunity for removing trials with the low quality before any analyses
starts is the important contribution in improving the efficiency of data processing.
Additionally, it was shown that one point calibration enhances results for the dis-
tance based algorithm. However, the improvement for non-calibrated samples does
not make this algorithm better than the regression based one. It shows that further
studies on more complicated regression based algorithms for using the eye movement
signal for human computer interaction may provide results improvement.
This is a pre-print. The final version of the paper was published in Springer, Smart Innovation,
Systems and Technologies, Vol. 39, 2015 as part of Proceedings of the 7th KES International
Conference on Intelligent Decision Technologies (KES-IDT 2015) and is available in Springer
Link library via http://dx.doi.org/10.1007/978-3-319-19857-6_31
6 Bibliography
1. Blignaut, Pieter; Wium, Daniël. The effect of mapping function on the accuracy of a vid-
eo-based eye tracker. In: Proceedings of the 2013 Con-ference on Eye Tracking South Af-
rica. ACM, 2013. p. 39-46
2. De Luca, A.; Denzel, M.; Hussmann, H.: Look into my eyes!: Can you guess my pass-
word?. In: Proceedings of the 5th Symposium on Usable Privacy and Security. ACM,
2009. p. 7.
3. Duchowski, Andrew. Eye tracking methodology: Theory and practice. Vol. 373. Springer,
2007.
4. Harężlak K., Kasprowski P., Stasch M.: Towards accurate eye tracker calibration meth-
ods and procedures. Proceedings of the 17th International Conference in Knowledge Based
and Intelligent Information and Engineering Systems - KES2014. Procedia Computer Sci-
ence, Vol. 35, pp. 1073 1081, Elsevier, 2014 (ScienceDirect)
5. Harężlak K., Kasprowski P.: Evaluating quality of dispersion based fixation detection al-
gorithm. 29th International Symposium on Computer and Information Sciences. Infor-
mation Sciences and Systems 2014. Springer Lecture Notes on Electrical Engineering, pp
97-104, 2014.
6. Holmqvist, Kenneth, et al. Eye tracking: A comprehensive guide to methods and measures.
Oxford University Press, 2011.
7. Jacob, Robert JK. "The use of eye movements in human-computer interaction techniques:
what you look at is what you get." ACM Transactions on Information Systems (TOIS) 9.2
(1991): 152-169.
8. Kasprowski P., Harężlak K., Stasch M.: Guidelines for the Eye Tracker Calibration Using
Points of Regard. Information Technologies in Biomedicine. Volume 4. Advances in Intel-
ligent Systems and Computing. Springer International Publishing, Vol. 284, pp. 225-236,
2014.
9. Kasprowski P., Harężlak K.,: Cheap and easy PIN entering using eye gaze. 3rd Conference
on Cryptography and Security Systems. Annales UMCS Informatica, Volume 14, 1, pp 75-
84, Lublin, 2014.
10. Kumar, M, et al.: Reducing shoulder-surfing by using gaze-based password entry. In: Pro-
ceedings of the 3rd symposium on Usable privacy and security. ACM, 2007. p. 13-19.
11. Ramanauskas, N.: Calibration of video-oculographical eye-tracking system. Electronics
and Electrical Engineering, 2006, 8.72: 65-68.
12. Salvucci, D. D., Goldberg, J. H.: Identifying fixations and saccades in eye-tracking proto-
cols. In Proceedings of the 2000 Symposiumon Eye Tracking Research and Applications,
7178, NY: ACM Press (2000)
13. Shic, F., Chawarska, K., Scassellati, B.: The incomplete fixation measure. In Proceedings
of the 2008 Symposium on Eye Tracking Research and Applications, 111114, NY: ACM
Press (2008)
... Such the method does not give accurate absolute gaze coordinates, but it may be used to check, if the eyes move at all. Our previous experiments have showed that when the eye tracker calibrated for one person is utilized for registering eye movements of other people, in most cases, the direction of a calculated gaze movement is similar, yet less accurate than for the calibrated user [20]. ...
... Of course, more points result in better calibration model, but even one point calibration may be valuable. For instance in [20], it has been shown that even after one point calibration, it was possible to evaluate at which of nine parts of a screen the user was looking. ...
Article
Full-text available
Vision plays a crucial role in children’s mental development. Therefore, early diagnosis of any vision disparities and implementation of a correct therapy is very important. However, carrying out such a procedure in case of young children and especially children with brain dysfunctions poses some limitations due to cooperation problems. The vision diagnostics and treatment (VisDaT) system presented in this paper is meant to help therapists in proper diagnosis and treatment involving such children. It utilizes a computer connected to two monitors and equipped with a specialized software. The main system components are as follows: an eye tracker recording child’s eye movements and a digital camera monitoring online child’s reactions. The system is equipped with a specialized software, which creates the opportunity to stimulate children’s vision with a dedicated stimulus and post hoc analyses of recorded sessions, which enable making decision as to the future treatment.
... It was observed during our previous experiments that, when an eye tracker is calibrated for one person, it is possible to utilize an output it produces when another person is using the same eye tracker. Obviously, the signal must be recalibrated to show true gaze points of this new person, however, even before this recalibration it is possible to see that eye is moving and determine the direction of this movement [8]. ...
... After collecting at least four such points, the application is able to calculate a calibration function and use it to show child's gaze point more accurately. Of course the accuracy increases when more points are available[8]. ...
Chapter
Children affected by brain disabilities require a lot of attention and care to improve their life. The quicker the support will be introduced the better effect can be achieved. One of important impairments resulted from the brain disability is a cerebral visual one. In case of children, communicating with whom is difficult or impossible, assessment of vision quality is very challenging and eye tracking methods may prove very useful. The research discussed in this paper was devoted to development of a workspace, which may support the effort of therapists working on improving the quality of disabled children's life. The important element of this solution is the implicit calibration procedure, making eye movement registration possible. Additionally, there were several stimuli developed and tested with cooperation of therapists from one of associations for children with developmental disabilities. Initial tests confirmed usefulness of the elaborated solution, which facilitates children's vision assessment based on the eye movement signal and may be used for a further children therapy.
... ACM ISBN 978-1-4503-7135-3/20/06. https://doi.org/10.1145/3379157.3391419 difficult to be used for practical solutions [Kasprowski and Harezlak 2017]. ...
Conference Paper
The purpose of the paper is to test the possibility of identifying people based on the input they provide to an eye tracker during the calibration process. The most popular eye trackers require the calibration before their first usage. The calibration model that is built can recalculate the subsequent eye tracker's output to genuine gaze points. It is well known that the model is idiosyncratic (individual for the person). The calibration should be repeated every time the person uses the eye tracker. However, there is evidence that the models created for the same persons may be reused by them (but obviously with some loss of accuracy). The general idea investigated in this paper is that if we take an uncalibrated eye tracker's output and compare it with the genuine gaze points, the errors will be repeatable for the same person. We tested this idea using three datasets with an eye tracker signal recorded for 52 users. The results are promising as the accuracy of identification (1 of N) for the datasets varied from 49% to 71%.
... The research presented in the paper aims at checking if it is possible to use an inexpensive remote eye tracker without any calibration as an input controller for games [Kasprowski and Harezlak 2017]. Section 2 describes an experiment we conducted with 66 users, during which we registered the uncalibrated signal together with the ground-truth information about the actual gaze coordinates. ...
Conference Paper
It seems that controlling games with the eyes should be very intuitive and obvious. However, eye-controlled games have not become very popular yet. One of the reasons is - in our opinion - the necessity of eye tracker calibration before its every usage. This process is not very long, but it is inconvenient and requires focusing on the particular task. Moreover, sometimes the calibration fails and must be repeated. According to our observations, even when the eye tracker is not calibrated for the specific user, there is some information in the registered signal that may be used to control a game. Of course, without the calibration, the eye tracker signal lacks accuracy and precision. However, it is acceptable for some types of games. The main contribution of the paper is checking to what extent an uncalibrated eye tracker signal may be used in a gaming environment. At first, a simple experiment was prepared to verify if the gaze location and eye movement direction may be estimated, having only the uncalibrated signal. Then the idea was tested in the field study involving several hundred participants.
... It may be assumed that although the ET signal does not reflect the exact gaze location, it corresponds to it [Kasprowski and Harezlak 2017]. So, it is possible to build a function (called the calibration model) that recalculates the ET signal into the gaze location -if some real gaze locations are known. ...
Conference Paper
Eye movement-based biometric has been developed for over 15 years, but for now - to the authors' knowledge - no commercial applications utilize this modality. There are many reasons for this, starting from still low accuracy and ending with the problematic setup. One of the essential elements of this setup is the calibration , as nearly every eye tracker needs to be calibrated before its first usage. This procedure makes any authentication based on eye movement a cumbersome and lengthy process. The main idea of the research presented in this paper is to perform authentication based on a signal from a cheap remote eye tracker but - contrary to the previous studies - without any calibration of the device. The uncalibrated signal obtained from the eye tracker is used directly, which significantly simplifies the enrollment process. The experiment presented in the paper aims at protection from a so-called "lunchtime attack" when an unauthorized person starts using a computer, taking advantage of the absence of the legitimate user. We show that such an impostor may be detected with an analysis of the signal obtained from the eye tracker when the user clicks with a mouse objects on a screen. The method utilizes the assumptions that: (1) users usually look at the point they click, and (2) an uncalibrated eye tracker signal is different for different users. It has been shown that after the analysis of nine subsequent clicks, the method is able to achieve the Equal Error Rate lower than 15% and may be treated as a valuable and difficult to counterfeit supplement to classic face recognition and password-based computer protection methods.
... Among the existing commercial devices, Eye Tribe (The Eye Tribe © , The Eye Tribe, Copenhagen, Denmark) has already been used in several research studies and applications (e.g., References [28][29][30][31]). Based on the results of a recent study, presented by [26], the accuracy and the precision of this device may be considered comparable with these of well-established trackers. ...
Article
Full-text available
The present study evaluates the quality of gaze data produced by a low-cost eye tracker (The Eye Tribe©, The Eye Tribe, Copenhagen, Denmark) in order to verify its suitability for the performance of scientific research. An integrated methodological framework, based on artificial eye measurements and human eye tracking data, is proposed towards the implementation of the experimental process. The obtained results are used to remove the modeled noise through manual filtering and when detecting samples (fixations). The outcomes aim to serve as a robust reference for the verification of the validity of low-cost solutions, as well as a guide for the selection of appropriate fixation parameters towards the analysis of experimental data based on the used low-cost device. The results show higher deviation values for the real test persons in comparison to the artificial eyes, but these are still acceptable to be used in a scientific setting.
... These low-cost devices are basically lower in resolution and calibration phase plays a major role in determining the quality of the data. Multiple calibrations seem an attractive means but accomplishing it, is often cumbersome and exhaustive process [10]. Bereft of the modes of the calibration, the inherent noise namely, variable and systematic, poses major challenge for using eye tracker data. ...
Article
Full-text available
Eye tracking is one of the most widely used technique for assessment, screening and human-machine interaction related applications. There are certain issues which limit the usage of eye trackers in practical scenarios, viz., i) need to perform multiple calibrations and ii) presence of inherent noise in the recorded data. To address these issues, we have proposed a protocol for one-time calibration against the “regular” or the “multiple” calibration phases. It is seen that though it is always desirable to perform multiple calibration, the one-time calibration also produces comparable results and might be better for individuals who are not able to perform multiple calibrations. In that case, “One-time calibration” can also be done by a participant and the calibration results are used for the rest of the participants, provided the chin rest and the eye tracker positions are unaltered. The second major issue is the presence of the inherent noise in the raw gaze data, leading to systematic and variable errors. We have proposed a signal processing chain to remove these two types of errors. Two different psychological stimuli-based tasks, namely, recall-recognition test and number gazing task are used as a case study for the same. It is seen that the proposed approach gives satisfactory results even with one-time calibration. The study is also extended to test the effect of long duration task on the performance of the proposed algorithm and the results confirm that the proposed methods work well in such scenarios too.
... Authors of [12][13] discuss some interesting applications of ribbon menus, while [14] gives a review of sophisticated interfaces of medical devices. Examples for alternative, adaptable interfaces and interactions designed to support disabled persons are given in [15] [16][17] [18] and [19]. Novel methods of interaction design for multimedia ap- plications and computer games are discussed in [20] and [21]. ...
... A classification model was built based on N training samples, with usage of an SVM classifier [48]. Using data of a similar structure utilized in our previous research [49] and a grid search algorithm, we obtained the best results for the RBF kernel with gamma ¼ 2 À9 and C ¼ 2 15 . Therefore, these values were used in the current research. ...
Article
Full-text available
This paper presents the first attempt to fuse two different kinds of behavioral biometrics: mouse dynamics and eye movement biometrics. Mouse dynamics were collected without any special equipment, while an affordable The Eye Tribe eye tracker was used to gather eye movement data at a frequency of 30 Hz, which is also potentially possible using a common web camera. We showed that a fusion of these techniques is quite natural and it is easy to prepare an experiment that collects both traits simultaneously. Moreover, the fusion of information from both signals gave 6.8 % equal error rate and 92.9 % accuracy for relatively short registration time (20 s on average). Achieving such results were possible using dissimilarity matrices based on dynamic time warping distance.
Conference Paper
Full-text available
With growing access to cheap low end eye trackers using simple web cameras, there is also a growing demand on easy and fast usage of this devices by untrained and unsupervised end users. For such users the necessity to calibrate the eye tracker prior to its first usage is often perceived as obtrusive and inconvenient. In the same time perfect accuracy is not necessary for many commercial applications. Therefore, the idea of implicit calibration attracts more and more attention. Algorithms for implicit calibration are able to calibrate the device without any active collaboration with users. Especially, a real time implicit calibration, that is able to calibrate a device on-the-fly, while a person uses an eye tracker, seems to be a reasonable solution to the aforementioned problems. The paper presents examples of implicit calibration algorithms (including their real time versions) based on the idea of probable fixation targets (PFT). The algorithms were tested during a free viewing experiment and compared to the state of the art PFT based algorithm and explicit calibration results.
Conference Paper
Full-text available
Information hidden in the eye movement signal can be a valuable source of knowledge about a human mind. This information is commonly used in multiple fields of interests like psychology, medicine, business, advertising or even software developing. The proper analysis of the eye movement signal requires its elements to be extracted. The most important ones are fixations-moments when eyes are almost stable and the brain is acquiring information about the scene. There were several algorithms, aiming at detecting fixations, developed. The studies presented in this paper focused one of the most common dispersion based algorithm-I-DT one. The various ways of evaluating its results were analyzed and compared. Some extensions in this algorithm were made as well.
Article
Full-text available
Eye movement is a new emerging modality in human computer interfaces. With better access to devices which are able to measure eye movements (so called eye trackers) it becomes accessible even in ordinary environments. However, the first problem that must be faced when working with eye movements is a correct mapping from an output of eye tracker to a gaze point – place where the user is looking at the screen. That is why the work must always be started with calibration of the device. The paper describes the process of calibration, analyses of the possible steps and ways how to simplify this process.
Conference Paper
Full-text available
PINs are one of the most popular methods to perform simple and fast user authentication. PIN stands for Personal Identification Number, which may have any number of digits or even letters. Nevertheless, 4-digit PIN is the most common and is used for instance in ATMs or cellular phones. The main ad-vantage of the PIN is that it is easy to remember and fast to enter. There are however some drawbacks. One of them – addressed in this paper – is a possibility to stole the PIN by a technique called 'shoulder surfing'. To avoid such problems a novel method of the PIN entering was proposed. Instead of using a numerical keyboard, the PIN may be entered by eye gazes, which is a hands-free, easy and robust technique.
Conference Paper
Full-text available
Eye movement data may be used for many various purposes. In most cases it is utilized to estimate a gaze point - that is a place where a person is looking at. Most devices registering eye movements, called eye trackers, return information about relative position of an eye, without information about a gaze point. To obtain this information, it is necessary to build a function that maps output from an eye tracker to horizontal and vertical coordinates of a gaze point. Usually eye movement is recorded when a user tracks a group of stimuli being a set of points displayed on a screen. The paper analyzes possible scenarios of such stimulus presentation and discuses an influence of usage of five different regression functions and two different head mounted eye trackers on the results.
Conference Paper
Full-text available
In a video-based eye tracker the pupil-glint vector changes as the eyes move. Using an appropriate model, the pupil-glint vector can be mapped to coordinates of the point of regard (PoR). Using a simple hardware configuration with one camera and one infrared source, the accuracy that can be achieved with various mapping models is compared with one another. No single model proved to be the best for all participants. It was also found that the arrangement and number of calibration targets has a significant effect on the accuracy that can be achieved with the said hardware configuration. A mapping model is proposed that provides reasonably good results for all participants provided that a calibration set with at least 8 targets is used. It was shown that although a large number of calibration targets (18) provide slightly better accuracy than a smaller number of targets (8), the improvement might not be worth the extra effort during a calibration session.
Article
Full-text available
Holmqvist, K., Nyström, N., Andersson, R., Dewhurst, R., Jarodzka, H., & Van de Weijer, J. (Eds.) (2011). Eye tracking: a comprehensive guide to methods and measures, Oxford, UK: Oxford University Press.
Conference Paper
Full-text available
Authentication systems for public terminals { and thus pub- lic spaces { have to be fast, easy and secure. Security is of utmost importance since the public setting allows mani- fold attacks from simple shoulder surng to advanced ma- nipulations of the terminals. In this work, we present Eye- PassShapes, an eye tracking authentication method that has been designed to meet these requirements. Instead of using standard eye tracking input methods that require precise and expensive eye trackers, EyePassShapes uses eye ges- tures. This input method works well with data about the rel- ative eye movement, which is much easier to detect than the precise position of the user's gaze and works with cheaper hardware. Dierent evaluations on technical aspects, usabil- ity, security and memorability show that EyePassShapes can signicantly increase security while being easy to use and fast at the same time.
Conference Paper
Full-text available
The process of fixation identification—separating and labeling fixations and saccades in eye-tracking protocols—is an essential part of eye-movement data analysis and can have a dramatic impact on higher-level analyses. However, algorithms for performing fixation identification are often described informally and rarely compared in a meaningful way. In this paper we propose a taxonomy of fixation identification algorithms that classifies algorithms in terms of how they utilize spatial and temporal information in eye-tracking protocols. Using this taxonomy, we describe five algorithms that are representative of different classes in the taxonomy and are based on commonly employed techniques. We then evaluate and compare these algorithms with respect to a number of qualitative characteristics. The results of these comparisons offer interesting implications for the use of the various algorithms in future work.
Conference Paper
Shoulder-surfing - using direct observation techniques, such as looking over someone's shoulder, to get passwords, PINs and other sensitive personal information - is a problem that has been difficult to overcome. When a user enters information using a keyboard, mouse, touch screen or any traditional input device, a malicious observer may be able to acquire the user's password credentials. We present EyePassword, a system that mitigates the issues of shoulder surfing via a novel approach to user input. With EyePassword, a user enters sensitive input (password, PIN, etc.) by selecting from an on-screen keyboard using only the orientation of their pupils (i.e. the position of their gaze on screen), making eavesdropping by a malicious observer largely impractical. We present a number of design choices and discuss their effect on usability and security. We conducted user studies to evaluate the speed, accuracy and user acceptance of our approach. Our results demonstrate that gaze-based password entry requires marginal additional time over using a keyboard, error rates are similar to those of using a keyboard and subjects preferred the gaze-based password entry approach over traditional methods.