CREATING AUDIENCE SPECIFIC GALACTIC SIMULATIONS

USING EYE-TRACKING TECHNOLOGY

Janelle Arita

Department of Computer Science, Depauw University, 313 S. Locust Street, Greencastle, Indiana, U.S.A.

Jenniffer Feliz

Department of Computer Science, Fordham University, 113 West 60th Street, New York City, New York, U.S.A.

Dennis Rodriguez, Hans-Peter Bischof, Manjeet Rege, Reynold Bailey

Department of Computer Science, Rochester Institute of Technology, 1 Lomb Memorial Drive, Rochester, New York, U.S.A.

Keywords:

Perception, Eye-tracking, Data visualization, Galactic images.

Abstract:

This study uses eye-tracking to investigate the differences in the way professional astrophysicists and novices

observe simulations of galactic events. The results of this study provide insight into which aspects of the data

are important and allow us to tailor the visualizations for a speciﬁc group. We hypothesized that the gaze

patterns of professionals and novices would vary considerably. A user study was performed on two groups:

trained astrophysicists and novices. Each group was presented with a randomized sequence of images and a

video while their gaze patterns were recorded with an eye-tracker. We discovered that although both groups

observed each image for the same duration, experts limited their ﬁxations to a smaller area. Novices, on the

other hand, had ﬁxations which were spread across the images. For the video, the astrophysicists were more

focused on simulations in which most of the data was visible and the camera angles had minimal axis change.

1 INTRODUCTION

Data visualization techniques allow us to better un-

derstand large multidimensional data and often re-

veal details that would generally be overlooked. The

Center for Computational Relativity and Gravitation

(CCRG) at Rochester Institute of Technology aims to

further our understanding of astrophysical phenom-

ena through the use of mathematical modeling, super-

computing and data visualization (Rochester Institute

of Technology, 2010). The faculty at CCRG have de-

veloped an extensible and robust visualization system

named Spiegel. Spiegel is highly versatile and capa-

ble of visualizing large amounts of multidimensional

data, which allows for the analysis of the data in both

time and space (Bischof, 2010). The Spiegel sytem

has been shown to be useful for a wide range of visu-

alization applications (Bischof et al., 2006) and sev-

eral images generated by Spiegel have been featured

in documentaries about black holes (A&E Television

Networks, 2010).

From a visualization perspective, it is crucial to un-

derstand whether the simulations created are scientif-

ically meaningful and relevant to the professionals in

the ﬁeld. In this paper, we present an eye-tracking

study that investigates the differences in the way that

experts and novices look at galactic simulations. By

extracting ﬁxations from the eye-tracking data, we

are able to determine the areas of interest in the data

for each group. Understanding what aspects of the

data the astrophysicists consider important allows for

the creation of customized simulations which will be

more relevant to professionals. The beneﬁt of captur-

ing eye movement data, as opposed to simply asking

what viewers are paying attention to, is that people

are often not aware of their own behavioral strate-

gies (Eger et al., 2007). With eye-tracking data, vis-

ibility, meaningfulness, and placement can be objec-

tively evaluated and the ﬁndings can be used to im-

prove design (Goldberg and Kotval, 1999).

We hypothesized that experts would look at the

simulations very differently than novices. It is prob-

218

Arita J., Feliz J., Rodriguez D., Bischof H., Rege M. and Bailey R..

CREATING AUDIENCE SPECIFIC GALACTIC SIMULATIONS USING EYE-TRACKING TECHNOLOGY.

DOI: 10.5220/0003321402180223

In Proceedings of the International Conference on Imaging Theory and Applications and International Conference on Information Visualization Theory

and Applications (IVAPP-2011), pages 218-223

ISBN: 978-989-8425-46-1

 2011 SCITEPRESS (Science and Technology Publications, Lda.)

able that novices will focus more on visually appeal-

ing parts of the simulation or image, while experts

will focus on areas that they ﬁnd more important be-

cause of their advanced knowledge. In order to inves-

tigate this, our participants were presented with im-

agery from the Hubble Space Telescope (Space Tele-

scope Science Institute, 2010) as well as images and

a video generated by the Spiegel system.

2 RELATED WORK

Eye-tracking systems ﬁrst emerged in the early

1900s (Dodge and Cline, 1901). More recently, eye-

tracking has been used within a variety of contexts

including aviation, driving, market research, psychol-

ogy, and design (Goldberg and Kotval, 1999). Eye-

trackers have also been used for human-computer in-

teraction (Hutchinson et al., 1989), interactive graph-

ics applications (Luebke et al., 2002; Levoy and

Whitaker, 1990; O’Sullivan et al., 2003), and large

scale display systems (Baudisch et al., 2003).

Researchers have investigated eye movements of

novices and experts in different scenarios including

surgeons performing aiming tasks on a laparoscopic

surgery simulator (Law et al., 2004). A method has

also been presented for improving visualization sys-

tems by using eye-tracking to determine what regions

of the data viewers considered to be important (Lu

et al., 2010). Based on this information, reason-

able rendering parameters were computed and the in-

teractivity of the visualization system improved sig-

niﬁcantly. Grindinger and Duchowski automatically

classiﬁed novice and expert pilots by analyzing their

gaze patterns (Grindinger et al., 2010). The aim of

our study is to build on the existing novice versus ex-

pert research, and to obtain the advantages that eye-

tracking can offer for improving data visualization.

3 EXPERIMENTAL DESIGN

3.1 Apparatus

A Mirametrix S1 eye-tracker operating at 60 Hz with

gaze position accuracy less than 1 degree was used.

The eye-tracker uses infrared illumination and an in-

frared camera to record video of the observer’s eyes.

The video is analyzed to locate the corneal reﬂection

and pupil center and this information is used to de-

termine the location on the screen where the observer

is looking (Poole and Ball, 2005). Stimuli were pre-

sented on a 20 inch LCD monitor with a resolution

of 1680x1050. Figure 1 shows a photograph of the

experiment setup. The stimuli consisted of sixteen

images (10 from Hubble Space Telescope, 6 from

Spiegel) as well as a video generated using Spiegel.

The images used are shown in Figure 2.

Figure 1: Experimental setup.

3.2 Participants

Twenty participants between the ages of 18 to 36 vol-

unteered for this study. All participants were screened

for impairments in visual acuity and color vision. Par-

ticipants were separated into two groups - novices and

experts, based on their knowledge of astrophysics.

Sixteen participants were grouped into the novice cat-

egory while 4 were grouped as experts.

• Expert Group. Participants were selected for the

expert group if they held at least a Masters degree

in astrophysics. The selection of participants in

the expert group was limited due to the criteria of

knowledge on astrophysics. There were 4 male

participants with an average age of 34. Partici-

pants in this group were asked to rate their knowl-

edge of astrophysics on a scale from 1 to 10 with

10 being the highest. The average rating was 7.25.

• Novice Group. Participants were selected for the

novice group if they had little or no knowledge

about astrophysics. Knowledge was determined

based on participants’ prior courses and hobbies.

There were 10 females, and 7 males with an av-

erage age of 21. The average rating of knowledge

on astrophysics for the novice group was 2.58.

3.3 Procedure

The user study took place in a controlled room with

moderate lighting. Each participant was seated ap-

proximately 20 inches in front of the screen and po-

sitioned such that both pupils were clearly identiﬁed

CREATING AUDIENCE SPECIFIC GALACTIC SIMULATIONS USING EYE-TRACKING TECHNOLOGY

219

Figure 2: The 16 images that were shown as stimuli for the ﬁrst phase of the study.

by the eye-tracker. Participants were asked to mini-

mize head movement during the experiment to ensure

accurate eye tracking.

The participants were asked to view a randomized

sequence of the sixteen images followed by the video.

They were instructed to rate the quality of each image

on a scale from 1 to 10 with 10 being the highest.

Following each image, a blank screen was displayed

for 3 seconds. While the blank screen was being dis-

played, the participants verbally expressed their rating

for the image they just viewed. The participants were

not given a deﬁnition of quality. Instead it was left

up to them to formulate their own notion of quality.

The purpose of the ratings was to help the participant

focus on the areas of the visuals which they consid-

ered most appealing. The complete set of instructions

read verbatim to each participant can be found in Ap-

pendix A. Eye-tracking data collected from each par-

ticipant was saved into an XML ﬁle. Each XML ﬁle

contained the following data: time elapsed, location

of ﬁxation for both left and right eye, ﬁxation dura-

tion, ﬁxation ID, and ﬁle name of the image or video

being displayed. Testing for each participant, includ-

ing calibration, and vision and color tests, lasted ap-

proximately 10 minutes.

4 RESULTS AND DISCUSSION

4.1 Images

4.1.1 Percentage of Fixation Duration

To determine if the gaze patterns for the novices and

experts were different, we divided each image into an

8 x 8 grid and compared the number of ﬁxations in

corresponding grid cells for the experts and novices.

Figure 3 shows the average ﬁxation distribution for

the group of novices for one of the images in the test

set. Figure 4 shows the average ﬁxation distribution

for the group of experts for the same image. Notice

that the gaze pattern for the novices tends to be more

spread out while the gaze pattern for the experts is

more focused. A similar trend was observed for all

of the images in the experiment. Although the gaze

patterns for the experts tended to be more focused

while the gaze patterns for the novices tended to be

more spread out, there was still a strong correlation

between the two groups. The Pearson coefﬁcient of

correlation between experts and novices for all six-

teen images was r = 0.8763. This suggests that both

groups still spent the majority of time attending to the

same salient features and likewise ignoring the same

non-salient features. This supports the ﬁndings from

previous studies which show that our focus is natu-

rally drawn to regions of bright saturated colors, high

IVAPP 2011 - International Conference on Information Visualization Theory and Applications

220

edge density, high local contrast, and informative re-

gions of an image (Mackworth and Morandi, 1967;

Mannan et al., 1996; Parkhurst and Niebur, 2003).

Figure 3: Average ﬁxation time for novices for one image

from the test set.

Figure 4: Average ﬁxation time for experts for one image

from the test set.

4.1.2 Number of Fixations

The average number of ﬁxations per image was also

calculated for both groups (see Figure 5). More ﬁxa-

tions per image may be an indicator or higher interest

or focus or the result of a rapid visual search (Gold-

berg and Kotval, 1999). The number of ﬁxations

for the Hubble images tended to be higher than the

number of ﬁxations for the Spiegel images. This is

likely due to the fact that the Hubble photographs had

richer content across the entire image compared to

the Spiegel images. We also observed that there was

a strong correlation between the number of ﬁxations

and the average quality rating. In particular, a larger

number of ﬁxations tended to result in a higher quality

rating. The Pearson coefﬁcient of correlation between

image ratings and the number of ﬁxations for experts

was r = 0.7965 and for novices it was r = 0.8750.

4.2 Video

One of the problems that is often faced when creat-

ing visualizations is determining the best path that the

Figure 5: Average number of ﬁxations for novices and ex-

perts.

Figure 6: Percentage of ﬁxation time spent in each quadrant

of the video for the novices.

Figure 7: Percentage of ﬁxation time spent in each quadrant

of the video for the experts.

camera takes through the data. Although very little

work has been done on interpreting eye-tracking data

of dynamic stimuli we believe that it holds great po-

tential for aiding the design process. In the second

phase of our study, we displayed a video to the par-

ticipants. The video was divided into four quadrants,

CREATING AUDIENCE SPECIFIC GALACTIC SIMULATIONS USING EYE-TRACKING TECHNOLOGY

221

Figure 8: Screen shot of preferred ﬂight path for segments 3 to 7.

each showing a different ﬂight path through the same

data. The video in each quadrant was synchronized so

that each event in the simulation occurred at the same

time from different perspectives.

By displaying four ﬂight paths at once, we deter-

mined which ﬂight path was attended to most often by

the two groups of participants. We determined this by

calculating the ﬁxation duration for each ﬂight path

for each group. The ﬂight path that contained the most

ﬁxation duration was the ﬂight path that was most ob-

served. Using this approach, we determined that the

experts spent more time looking at the ﬂight path in

the lower left quadrant, while the novices preferred

the ﬂight path in the top left. See Figures 6 and 7.

This observation by itself is not very useful unless

we take it one step further and analyze which of the

four quadrants was attended to at speciﬁc stages of

playback. To accomplish this, we divided the video

into 14 segments, each segment spanning 3 seconds in

length. By analyzing the video in smaller segments,

we distinguished which ﬂight path a participant ﬁx-

ated on at speciﬁc times. For example, Figure 8 shows

a screen shot of the preferred ﬂight path for segments

3 to 7. We recreated an optimal audience speciﬁc

video for each group by merging the preferred ﬂight

path segments into a single video.

5 CONCLUSIONS AND FUTURE

WORK

We have presented an experiment which used eye-

tracking to compare the gaze patterns of novices and

expert astrophysicists while looking at various im-

ages and simulations of galactic events. We noted

that although experts and novices both focus on sim-

ilar salient regions of an image and ignore similar

non-salient regions that the novices’ gaze tended to

be more spread out while the experts’ gaze tended

to be restricted to fewer regions of the image. For

the video, it was observed that the professional group

was more focused on simulations in which most of the

data was visible and the camera angles had minimal

axis change. These results provide valuable insight

that will be used for designing more relevant, audi-

ence speciﬁc visualizations.

Future work includes improvement of the simu-

lation video by creating smoother camera transitions

between segments. Instead of manually creating a

simulation by merging segments, we can recreate the

entire video by simulating the preferred camera ﬂight

path through the Spiegel simulation software. Ideally

we would like to completely automate this process to

allow the user to view several ﬂight paths and have the

system dynamically create the ﬁnal simulation video

based on the viewer’s gaze preferences. In order to

accomplish this, it will be necessary to interface the

eye-tracker with the simulation software.

REFERENCES

A&E Television Networks (2010). Cosmic holes: The

History Channel. http://www.history.com/shows/the-

universe/episodes/season-2.

Baudisch, P., DeCarlo, D., Duchowski, A., and Geisler, W.

(2003). Focusing on the essential: considering atten-

tion in display design. Commun. ACM, 46(3):60–66.

Bischof, H.-P. (2010). The Spiegel Project.

http://spiegel.cs.rit.edu/∼hpb/grapecluster/Spiegel/

index.html.

Bischof, H.-P., Dale, E., and Peterson, T. (2006). Spiegel - a

visualization framework for large and small scale sys-

tems. In In MSV 06: Proceedings of the 2006 Interna-

tional Conference of Modeling Simulation and Visual-

ization Methods.

IVAPP 2011 - International Conference on Information Visualization Theory and Applications

222

Dodge, R. and Cline, T. S. (1901). The angle velocity of

eye movements. Psychological Review, 8:145–157.

Eger, N., Ball, L. J., Stevens, R., and Dodd, J. (2007).

Cueing retrospective verbal reports in usability testing

through eye-movement replay. In BCS-HCI ’07: Pro-

ceedings of the 21st British HCI Group Annual Con-

ference on People and Computers, pages 129–137,

Swinton, UK, UK. British Computer Society.

Goldberg, J. H. and Kotval, X. P. (1999). Computer inter-

face evaluation using eye movements: methods and

constructs. International Journal of Industrial Er-

gonomics, 24(6):631 – 645.

Grindinger, T., Duchowski, A. T., and Sawyer, M. (2010).

Group-wise similarity and classiﬁcation of aggregate

scanpaths. In ETRA ’10: Proceedings of the 2010

Symposium on Eye-Tracking Research & Appli-

cations, pages 101–104, New York, NY, USA. ACM.

Hutchinson, T. E., White, K. P., Martin, W. N., Reichert,

K. C., and Frey, L. A. (1989). Human-computer in-

teraction using eye-gaze input. Systems, Man and Cy-

bernetics, IEEE Transactions on, 19(6):1527–1534.

Law, B., Atkins, M. S., Kirkpatrick, A. E., and Lomax,

A. J. (2004). Eye gaze patterns differentiate novice

and experts in a virtual laparoscopic surgery training

environment. In ETRA ’04: Proceedings of the 2004

symposium on Eye tracking research & applications,

pages 41–48, New York, NY, USA. ACM.

Levoy, M. and Whitaker, R. (1990). Gaze-directed volume

rendering. SIGGRAPH Comput. Graph., 24(2):217–

223.

Lu, A., Maciejewski, R., and Ebert, D. S. (2010). Volume

composition and evaluation using eye-tracking data.

ACM Trans. Appl. Percept., 7(1):1–20.

Luebke, D., Watson, B., Cohen, J. D., Reddy, M., and

Varshney, A. (2002). Level of Detail for 3D Graph-

ics. Elsevier Science Inc., New York, NY, USA.

Mackworth, N. H. and Morandi, A. J. (1967). The gaze

selects informative details within pictures. Perception

and Psychophysics, 2:547–552.

Mannan, S. K., Ruddock, K. H., and Wooding, D. S. (1996).

The relationship between the locations of spatial fea-

tures and those of ﬁxations made during visual exam-

ination of brieﬂy presented images. Spatial Vision,

10:165–188.

O’Sullivan, C., Dingliana, J., and Howlett, S. (2003). Eye-

movements and interactive graphics. The Mind’s

Eyes: Cognitive and Applied Aspects of Eye Move-

ment Research, pages 555–571. J. Hyona, R. Radach,

and H. Deubel (Eds.).

Parkhurst, D. and Niebur, E. (2003). Scene content selected

by active vision. Spatial Vision, 16:125–154.

Poole, A. and Ball, L. J. (2005). Eye tracking in human-

computer interaction and usability research: Current

status and future. In Prospects, Chapter in C. Ghaoui

(Ed.): Encyclopedia of Human-Computer Interaction.

Pennsylvania: Idea Group, Inc.

Rochester Institute of Technology (2010). Cen-

ter for Computatonal Relativity and Gravitation.

http://ccrg.rit.edu/.

Space Telescope Science Institute (2010). Hubblesite.

http://hubblesite.org.

APPENDIX

Instructions to Participants. In addition to the fol-

lowing instructions, which were read to the partic-

ipants at the start of the study, participants were

also given documentation showing that the study was

reviewed and approved by the Institutional Review

Board (IRB) at the institution where this study was

conducted.

The purpose of this study is to gain a better

understanding of how humans look at images

of galactic events.

You will be shown a sequence of images

and a video. Your task will be to evaluate the

quality of each image and video by assigning

a rating from 1 to 10 with 1 being the low-

est quality and 10 being the highest. Please

state your rating when the blank screen be-

tween images is displayed. The images that

you will be viewing consist of several images

from the Hubble Space Telescope as well as

computer generated images. You will also see

a computer generated video.

During the course of the experiment, a

noninvasive camera will be used to record

your eye movements. Please try to minimize

your head movements as this may adversely

affect the quality of the results. A short cal-

ibration process is necessary to ensure that

your eyes are being accurately tracked. This

will occur at the start of the experiment. Cal-

ibration simply involves looking at the targets

on the screen until they disappear. The entire

experiment should take no longer than 10 min-

utes to complete.

The results of this study may be pub-

lished in scientiﬁc research journals or pre-

sented at professional conferences. However,

your name and identity will not be revealed

and your record will remain anonymous. Your

name will not be used in any data collection,

so it will be impossible to tell your answers

from other peoples answers.

The potential beneﬁts of this study to so-

ciety include improvements in data visualiza-

tion techniques and the advancement of scien-

tiﬁc knowledge of human visual perception.

Participation is entirely voluntary. Addition-

ally, you may choose to withdraw from this

study at any time. If you decide not to par-

ticipate or to withdraw from this study, there

will not be a penalty to you. Do you have any

questions before we begin?

CREATING AUDIENCE SPECIFIC GALACTIC SIMULATIONS USING EYE-TRACKING TECHNOLOGY

223