=Paper=
{{Paper
|id=Vol-2470/p33
|storemode=property
|title=Accuracy of throwing distance perception in virtual reality
|pdfUrl=https://ceur-ws.org/Vol-2470/p33.pdf
|volume=Vol-2470
|authors=Karolis Butkus,Tautvydas Čeponis
|dblpUrl=https://dblp.org/rec/conf/ivus/ButkusC19
}}
==Accuracy of throwing distance perception in virtual reality==
https://ceur-ws.org/Vol-2470/p33.pdf
Accuracy of throwing distance perception in Virtual
Reality
Karolis Butkus Tautvydas Čeponis
Department of Information Systems Department of Multimedia
Kaunas University of Technology Kaunas University of Technology
Kaunas, Lithuania Kaunas, Lithuania
e-mail: k.butkus@ktu.edu e-mail: tautvydas.ceponis@ktu.edu
Abstract — This article investigates how people perceive room prior to the experiment and half participating blindly.
distances in virtual reality (VR) and use that information to The first tested method for improved distance perception was
execute a representation of a real life throwing motion. In visual replication of a real world environment, the second
order to measure accuracy, this research proposes a throwing was walking interaction, which allowed participants to move
motion testing framework, which acquires metrics data from around the virtual environment prior to testing. The results
both the real and virtual environments. The results show, that concluded that walking interaction significantly increased the
the examinees tend to throw more accurately at longer accuracy of distance perception and size perception to a
distances and use excessive amounts of force. lesser degree. Furthermore, it was more effective than visual
replication in both scenarios.
Keywords—virtual reality, perception, accuracy, throwing
motion A similar study to research [3] was carried out at
Clemson University in 2011 [4]. In this experiment,
I. INTRODUCTION researchers investigated near-field egocentric distance
During the last decade virtual reality technology has estimations in an Immersive Virtual Environment and
significantly improved and is used in different technological compared it to real world distances. The experiment
spheres. The visual representation is becoming more realistic examined two methods: verbal and reach measurements.
and looks more natural. Although technology is evolving, it Participants had to report distances verbally and then show it
is hard to replicate human senses. Therefore, this study tries with their reach. Results show that both verbal and reach
to analyze how accurately people perceive virtual world methods tend to underestimate distance and that with an
distances when executing a throw. increase in distance deviation also increased. Another
interesting fact was that the verbal method was less accurate
This study presents a throwing motion testing framework
than the reach method.
to determine the differences between the virtual and real
world’s environment perception capabilities. It will discuss The study [5] made by three researchers from the
similar studies in the field related to perception and motion Dresden University of Technology attempted to find out
tracking, explain the testing framework and methodology, what factors mostly affect people’s estimations for distance
the experiment’s process, discussion about the results and in the virtual world. They arranged the factors in four
drawbacks of this study and the conclusion, possible future. groups: measurement methods, technical, compositional and
human factors. The research concluded that people tend to
A similar project [1] to determine the perception of underestimate distance and that to improve human distance
virtual reality was carried out in 2008 by researchers from
recognition skills - a rich, detailed environment and powerful
Aachen, Germany. In their experiment, they asked 23
technical hardware must be ensured. Such as high quality
participants to estimate distances to virtual reality objects in
graphics, carefully adjusted camera settings and virtual
three different environments. Results show that people tend
environment with a regularly structured ground texture.
to underestimate distances and that visual surroundings did
not affect results considerably. As mentioned a few times in other researches people tend
to underestimate distances in virtual reality and according to
Another article checked people’s ability to locate
Steven M. LaValle, the cause for that could be different gaps
themselves in a virtual environment. Their task was to point
between pupils [6]. If pupils in the real world are closer than
at themselves in a VR platform using a pointer. The
in the virtual world, the virtual environment looks larger to
experiments results stated that participants most commonly
the user and the other way round if the pupils are further
locate themselves at the upper region of their face and that
apart in the real world.
draws a conclusion that people in a virtual environment are
more head-centered. [2] II. THROWING MOTION TESTING FRAMEWORK
A more recent study [3] was carried out by researchers To determine the differences in perception between
from Iowa State University. The group examined prior reality and a virtual environment, we focused on the different
attempts at improving distance perception in a Virtual aspects of throwing kinematics in reality and VR. Three
environment (VE) and proposed a more thorough main characteristics are taken into consideration: throwing
methodology to measure the results by isolating unaccounted distance in reality, throwing distance in virtual reality and the
variables in past studies. The experiment tested the initial velocity of the hand tracker in a throwing motion.
participant’s size and distance perception in a VE replica of a
real world room with half of the examinees having seen the
© 2019 for this paper by its authors. Use permitted under Creative
Commons License Attribution 4.0 International (CC BY 4.0) 121
To measure the above mentioned features a throwing To collect quite accurate motion data the tracker is
simulation framework was created. During the testing attached to the palm of the participant and the ball is put on
procedure participants throw a 10 gram ball to three different top of the device (Fig. 3). When the person executes a throw
distances (2 meters, 3 meters, 4 meters) and a tracker the tracker captures the initial velocity, and upon slowing
attached to their hand transmits VR data which is recorded down the system initiates a throw in virtual reality and sends
digitally, while real life distance is measured with a ruler. the collected speed and data about the ball’s collision with a
Each participant has three attempts at three distances with the ground surface to a text file. The real distance is measured
virtual reality headset being used and another with it with a ruler and all collected digital and non-digital data is
mounted on top of their head for tracking accuracy. saved in a spreadsheet.
The testing system is developed using Unity Engine and
HTC Vive Pro VR headset and tracker. The framework’s
visual environment is a replica of the room where the
simulation was performed so it would not cause distractions
to the participants. Distances at which the ball is thrown and
standing position are marked in both the real and virtual
environments (Fig. 1 and Fig. 2).
Fig. 3. Ball throw in the experiment
III. EXPERIMENT
The main goal of the experiment is to determine how
accurate is a human’s perception at determining distances
using a virtual throwing mechanism compared to a real
world throw.
The experiment participants were six people: 4 males and
2 females. The participants age ranged from 19 to 25 years
Fig. 1. VR testing platform (user view) (mean age 22.3), all of them were healthy and didn’t suffer
from VR sickness. At the beginning of the test, the
participants were given time to practice throwing in virtual
reality and get used to it. Then the examinees did three
consecutive throws at specified distances without a headset
and then they had three attempts with the virtual reality
device. This process was repeated three times at three
different shooting distances. During the experiment,
participants were not allowed to move from the starting
position. The collected distance and velocity data was saved
in a spreadsheet.
The experiment’s results are presented in Table I where
every user’s average thrown distance is shown in a
centimeters format. Results of shots with virtual reality
equipment and without it are separated and the total average
of each baseline distance is calculated.
Fig. 2. Real testing scene view
From Table I it is easy to see that people throw the ball
most accurately at a distance of 3 or 4 meters when using the
In the experiment, HTC Vive Pro virtual reality headset VR headset, whereas at the 2 meter mark there is a 10
and tracker are both connected to a personal computer with centimeters deviation. However, with unobstructed vision
Windows 10 operating system. The tracker’s data collecting people throw the ball more accurately at the first and third
Base stations 2.0 were placed at 5 meter distances from each distances and in this case there is about a 10 centimeters
other, at opposing corners of the room. The testing deviation from the second distance. This data shows that
framework was built in Unity 2018.3.5.f1 with an with an increase in distance people’s throws tend to become
implemented SteamVR plugin. The original plugin’s view for more accurate, whereas near distances are more difficult to
the ball throw scene was edited so that it would replicate the judge.
experiment room and the stock throw function was modified
so that it didn’t require any buttons to be pushed. The throw
in the system is initiated when the tracker is swung and the
velocity of the tracker starts to slow down after the constant
increase in velocity at the start of the throw. The simulated
environment replica consists of a 9 meter by 6 meter square
room with an open top. The layout is positioned at the exact
locations of real world objects.
122
Data about the average initial velocity is presented in a
clustered columns chart and a scatter graph (Fig. 5 and Fig.
TABLE I. AVERAGE THROWN DISTANCE WITH VR AND WITHOUT IT 6) where the baseline distances and different environments
are separated. Besides average values, medians are given to
USER
Reality VR make the data more accurate.
200 cm 300 cm 400 cm 200 cm 300 cm 400 cm From Figures 5 and 6 it is noticeable that people tend to
1 189.667 298.000 378.000 215.000 326.000 414.000 throw the ball with more power when they are in a virtual
environment than when they are in the real world. This
2 207.333 301.000 389.000 204.667 307.000 398.333
statement also is reaffirmed by the medians of all throws in
3 212.000 295.333 382.667 198.000 301.667 375.000 real and virtual worlds.
4 167.667 268.667 389.000 164.000 277.667 478.500
5 183.000 286.000 415.667 178.333 301.667 407.000
6 217.333 278.667 428.667 182.000 301.667 347.667
a
AVG 196.167 287.944 397.167 190.333 302.611 403.417
b
SD 17.565 11.458 18.464 17.222 14.090 40.216
a.
AVG – Average
b.
SD – Standard deviation
In addition, from the bar chart shown in Fig. 4, which
represents the average miss distance from a mark (negative
value if it is shorter than the baseline distance and positive if
the average value is greater), it is noticeable that the
experiment participants tend to underestimate distances and Fig. 5. Average initial velocity and medians
throw the ball at a shorter distance. Only two columns show
a slight ball overthrow and both belong to results achieved in
virtual reality .
Fig. 6. Average initial velocity linear regression
Fig. 4. Average distance from baseline mark
IV. DISCUSSION
Table II shows every participant’s standard deviation of This study was conducted to find out how accurately
three throws and average standard deviation which is about people perceive the virtual environment and decide what
17 centimeters. Therefore, it can be said that the experiment amount of power is needed to throw the ball. To achieve this
needs an increase in participants and throw attempts to make goal 6 participants took part in the experiment where they
the experiment’s data even more accurate. had to throw a ball at 3 distances with and without a VR
headset.
TABLE II. THE STANDARD DEVIATION OF EACH PARTICIPANT
THROWS
After all tests, the collected data shows that people’s
accuracy with VR tends to increase with an increase in
distance and that the average initial speed tends to be higher
USER
Reality VR
than pitching the ball without the headset. To explain the
200 cm 300 cm 400 cm 200 cm 300 cm 400 cm increase in velocity we could say that because people are
1 7.409 8.287 29.063 17.795 32.934 39.047 more head-centered [2] in a virtual environment, they sense
that distance is further than it actually is. Moreover, people
2 10.656 4.967 13.928 19.754 15.895 2.625
are more likely to underthrow than overthrow the ball in real
3 11.225 17.632 2.494 28.891 14.055 18.239 life and the increase in velocity when using VR allows their
4 8.807 24.253 14.900 12.832 14.055 3.500 shots to be more precise. But when people are throwing close
range shots the distances spread out and accuracy decreases.
5 30.342 4.546 13.888 4.989 28.170 0.816
These results show, that the described method can be
6 15.965 30.214 29.915 9.899 45.492 36.736
used to calibrate hand strength in Virtual Environment fields,
such as gaming [7], simulations [8], gesture recognition
123
systems [9]. The motion force a person outputs in a fully Simulation Game", Information, vol. 9, no. 12, p. 293, 2018.
immersed virtual system has to be decreased by 3 – 5 % to Available: 10.3390/info9120293.
assure that the user’s perception of his virtual strength [8] E. Danevičius, R. Maskeliūnas, R. Damaševičius, D. Połap and M.
Woźniak, "A Soft Body Physics Simulator with Computational
matches the real world results and compensates their depth Offloading to the Cloud", Information, vol. 9, no. 12, p. 318, 2018.
perception in a VE. Available: 10.3390/info9120318.
To acquire more accurate estimations we cannot forget [9] A. Vaitkevičius, M. Taroza, T. Blažauskas, R. Damaševičius, R.
Maskeliūnas and M. Woźniak, "Recognition of American Sign
that all velocity data is collected by a wireless tracker and the Language Gestures ina Virtual Reality Using Leap Motion", Applied
real ball that was put on the tracker could interfere with Sciences, vol. 9, no. 3, p. 445, 2019. Available: 10.3390/app9030445.
results and that could be a reason why the standard deviation [10] D. Połap, M. Woźniak, C. Napoli and E. Tramontana, "Real-time
for a few participant’s throws was so high. cloud-based game management system via cuckoo search algorithm"
International Journal of Electronics and Telecommunications, vol.4,
In addition, to help the person better comprehend the n. 61, p. 333-338, 2015.
depth of a virtual world during the experiment it could be
allowed for the participants to walk around the room as
shown in research [3] and not undertake the whole
experiment from a standing position while only having to
trust their vision.
Furthermore, it was brought to the examinees attention,
that to get more accurate results the participants had to do a
bigger backswing while performing the throwing motion to
get a more consistent velocity and more suitable throw
initialization timings.
V. CONCLUSION AND FUTURE WORKS
In this study, we concluded, that people perceives 2 – 4
meter distances nearly the same as in real life. Moreover,
people tend to use 3 to 5 % more power when throwing a
ball in virtual reality than in real life. However, the used
methodology needs improvement (some throws standard
deviation is as high as 45 centimeters) to eliminate
unnecessary factors, such as inaccuracy of manual real world
measurements and signal integrity loss from ball position
relative to the sensor. Furthermore, a larger pool of
participants is needed to achieve precise data averages and
calculations. There is also the possibility to attach a separate
sensor to the ball that is being thrown by the participants,
thus eliminating the need for real world measurements by
allowing us to compare the data between both throws
directly. Although the research is not perfect it has
considerable potential to be used as a calibration tool for
various virtual reality fields which involve hand motion and
arm strength.
REFERENCES
[1] C. Armbrüster, M. Wolter, T. Kuhlen, W. Spijkers and B. Fimm,
“Depth Perception in Virtual Reality: Distance Estimations in Peri-
and Extrapersonal Space”, CyberPsychology & Behavior, vol. 11, no.
1, pp. 9-15, 2008. Available: 10.1089/cpb.2007.9935
[2] A. H. V. D. Veer, A. J. T. Alsmith, M. R. Longo, H. Y. Wong, and B.
J. Mohler, “Where am I in virtual reality?,” Plos One, vol. 13, no. 10,
2018.
[3] J. W. Kelly, L. A. Cherep, B. Klesel, Z. D. Siegel, and S. George,
“Comparison of Two Methods for Improving Distance Perception in
Virtual Reality,” ACM Transactions on Applied Perception, vol. 15,
no. 2, pp. 1–11, Mar. 2018
[4] P. Napieralski et al., "Near-field distance perception in real and
virtual environments using both verbal and action responses", ACM
Transactions on Applied Perception, vol. 8, no. 3, pp. 1-19, 2011.
Available: 10.1145/2010325.2010328.
[5] R. Renner, B. Velichkovsky and J. Helmert, "The perception of
egocentric distances in virtual environments - A review", ACM
Computing Surveys, vol. 46, no. 2, pp. 1-40, 2013. Available:
10.1145/2543581.2543590.
[6] S. LaValle, Virtual Reality. Cambridge University Press, 2016, pp.
153-154. Available: http://vr.cs.uiuc.edu/vrbooka4.pdf
[7] R. Buzys, R. Maskeliūnas, R. Damaševičius, T. Sidekerskienė, M.
Woźniak and W. Wei, "Cloudification of Virtual Reality Gliding
124