Audience Behavior around Large
Interactive Cylindrical Screens
Gilbert Beyer
1
, Florian Alt
2
, Jörg Müller
3
, Albrecht Schmidt
4
, Karsten Isakovic
5
,
Stefan Klose
5
, Manuel Schiewe
5
, Ivo Haulsen
5
1
Ludwig-Maximilians-University Munich, Germany, gilbert.beyer@ifi.lmu.de
2
University of Duisburg-Essen, Germany, florian.alt@uni-due.de
3
Deutsche Telekom Laboratories, TU Berlin, Germany, joerg.mueller@tu-berlin.de
4
VIS, University of Stuttgart, Germany, albrecht.schmidt@acm.org
5
Fraunhofer FIRST, Berlin, Germany, name.surname@first.fraunhofer.de
ABSTRACT
Non-planar screens, such as columns, have been a popular
means for displaying information for a long time. In con-
trast to traditional displays their digital counterparts are
mainly flat and rectangular due to current technological
constraints. However, we envision bendable displays to be
available in the future, which will allow for creating new
forms of displays with new properties. In this paper we ex-
plore cylindrical displays as a possible form of such novel
public displays. We present a prototype and report on a user
study, comparing the influence of the display shape on user
behavior and user experience between flat and cylindrical
displays. The results indicate that people move more in the
vicinity of cylindrical displays and that there is no longer a
default position when it comes to interaction. As a result,
such displays are especially suitable to keep people in mo-
tion and to support gesture-like interaction.
Author Keywords
Cylindrical screens, digital columns, display formats, public
displays, interactive surfaces, non-planar screens
ACM Classification Keywords
H.5.2. Information interfaces and presentation: User Inter-
faces.
General Terms
Human Factors.
INTRODUCTION
As display technology progresses and digital displays be-
come cheaper, larger, and more robust, traditional displays
in public spaces are being replaced by their digital counter-
parts. These digital displays can provide, among other bene-
fits, interactivity, either by touch or by sensing the move-
ment of the audience. Due to the deployed display tech-
nologies, nowadays the majority of displays are flat,
rectangular, and framed. However, we learned from history,
that there are many successful forms of non-planar displays.
One popular form were columns. Freestanding columns
have the benefit of high visibility due to their concise and
elevated shape, and can also provide more screen real estate
on the same floor space. In addition, columns were exten-
sively available inside buildings for structural reasons. Fa-
mous examples for ancient cylindrical displays are Trajan’s
Column in Rome (see Figure 1) or columns in the Hathor
temple in Egypt. Even nowadays the most popular form of
non-planar displays are cylindrical screens, such as cylin-
drical bulletin-boards, inflatable columns used at events, or
street furniture columns used for cultural information, pub-
lic announcements and ads (Morris or Litfaß Columns, see
also Figure 1). With advances in technologies, bendable
displays will allow for turning nearly any surface into a dis-
play and hence allow for creating displays of almost arbi-
trary shape and size for no additional costs compared to flat
displays.
Figure 1: Examples of classical non-planar screens: (1) Tra-
jan’s Column with inscriptions in Rome (2) Replica of the first
Litfaß Column in Berlin (3) Modern Citylight Column.
We opted to investigate cylindrical displays as one possible
form of novel, arbitrary-shaped public displays. Hence we
present a prototype of an interactive cylindrical display and
Permission to make digital or hard copies of all or part of this work for
personal or classroom use is granted without fee provided that copies are
not made or distributed for profit or commercial advantage and that copies
bear this notice and the full citation on the first page. To copy otherwise,
or republish, to post on servers or to redistribute to lists, requires prior
specific permission and/or a fee.
CHI 2011, May 7–12, 2011, Vancouver, BC, Canada.
Copyright 2011 ACM 978-1-4503-0267-8/11/05...$5.00.
report on a user study comparing user behavior in front of
classical, flat displays and cylindrical displays showing the
same content (we use the terms cylindrical display and col-
umn interchangeably). Our contribution is threefold:
• We present an interactive cylindrical display, which
reacts to passers-by by adapting the content according
to their movements.
• We show that people in front of cylindrical displays
tend to move more and explore the content from a
wider range of standing positions and discuss how this
influences the way content should be designed.
• We discuss advantages and disadvantages of flat and
cylindrical displays based on the findings of our study.
RELATED WORK
In recent years, a variety of prototypes of non-planar dis-
plays have been developed. Most public deployments, how-
ever, still use planar, framed displays and their designs are
based on implicit assumptions that may not hold for non-
planar displays. While a number of studies of audience be-
havior for planar displays exist, their results may not gener-
alize to non-planar displays. Finally, the specific case of
cylindrical displays introduces the new category of semi-
framed displays.
Interactive and Public Displays
Inspired by early work such as Media Spaces [4] and the
Digital Desk [29], a variety of interactive and public dis-
plays have been proposed. The Plasma Poster Network
[7,8] was a network of touch enabled public displays in-
stalled in hallways supporting office communication. Simi-
lar to our approach, many of such displays used computer
vision to enable interaction from a distance. The seminal
work of Myron Krueger explored vision-based interaction
with large displays where the displays reacted to the users’
movements [15]. In ReflectiveSigns [22], a network of dis-
plays reacted implicitly to the audience’s viewing behavior.
The Intelligent Kiosk [6] was an early example of an inter-
active Avatar on a public display that reacted to passing
people. Similarly, the AmiQuin [26] was a virtual manne-
quin in a shop window that reacted to the audiences’ body
movements. Finally, Malik proposed vision-based interac-
tion techniques with multiple fingers for distant displays
[19].
Designs of Non-Planar Screens
Organic user interfaces [12] have been proposed as com-
puter interfaces that use non-planar displays for input and
output. Lin et al. presented the i-m-Tube, an interactive tu-
bular display [18]. The first version uses a single projector
and a convex mirror to create a back-projected cylindrical
display, which supports multi-touch. Benko et al. present
Sphere [1], a multi-touch enabled spherical display. They
discuss unique properties of spherical displays and multi-
touch interaction techniques for such displays. Some of the
properties for spheres are similar to properties of cylindrical
displays (like that the user can not see the whole display at
any time), while others are not (like a smooth transition be-
tween vertical and horizontal surfaces). Lee et al. [17] pre-
sented foldable displays like fans, newspapers, scrolls and
umbrellas. Their prototypes were implemented using pro-
jection, infrared markers and the Nintendo Wii remote
camera. They discuss affordances of different shapes and
possible interaction techniques. Volumetric displays appear
to show content inside the display volume. Grossmann et al.
focus on such displays’ unique properties and show how
they can be used for collaborative applications [9,10].
Regarding cylindrical displays, a number of commercial
designs exist. Many commercial cylindrical displays use
rotating LED’s (Kinoton, Dynascan), where the drive sys-
tem has to be well adjusted to get a jitter-free image. Other
technical solutions include static mechanical designs in-
cluding rings of LED modules (Barco) and projection-based
setups. In [2], Benko outlines challenges when designing
gestural interactions with non-flat surface computing inter-
faces derived from the development of three prototypes (a
gesture-enabled flat display, a sphere, and a dome). Chal-
lenges identified include walk-up-and-use functionality,
linking heterogeneous devices, usability from multiple di-
rections, and compelling applications.
Assumptions of Current Designs for Flat Displays
Most current deployments of interactive public displays use
planar, framed displays enabling interaction either through
touch or body gestures. The CityWall [24], for example,
was a large multi-touch display installed in downtown Hel-
sinki that supported browsing photo collections. Worlds of
Information [14] was an extension of the same system to
include touch interaction with 3D spheres of photos. Magi-
cal Mirrors [21] was a deployment in downtown Berlin
where passers-by could see their own mirror image on a
display and interact with virtual objects through body
movements. Such designs for flat displays usually start
from a number of implicit assumptions that need to be ques-
tioned for non-flat displays. Many designs assume that: 1)
people stop walking before they interact, 2) users can per-
ceive the content of the entire screen at any time, 3) users
can see what other people do when interacting with the dis-
play, 4) shoulders are usually parallel to the display, 5) the
position centrally in front of the display is preferred, 6) con-
tent is not distorted. For example, touch displays are obvi-
ously difficult to use while walking, and even for Magical
Mirrors a walking user would quickly leave the camera
view. For the CityWall, users could easily scale photos so
big that they hide the view for others, which would be a
problem if users cannot observe the effects of their actions
on other parts of the screen. Similarly, users started to play
‘soccer’ by throwing photos around, which would be very
different if one could not observe the actions of other users.
Vogel [28] used a parallel shoulder position as an indicator
of user interest, but it is not clear if this is valid for non-
planar displays. Also, the distortion of content could render
many current interfaces unusable on non-planar displays.
Audience Behavior in Front of Flat Displays
Any observations of audience behavior towards public dis-
plays have been conducted using planar displays, and it is
unclear whether existing findings generalize towards non-
planar displays. Scott et al. [27] explore how people use re-
gions on round tabletops, which share some properties with
cylindrical displays, but where the display itself is flat.
Huang et al. [13] show that for existing planar displays,
very few people look at them or stop in front of them. This
may also be related to the fact that many flat displays are
installed either on walls, orthogonal to walking direction, or
above eye height. Cylindrical screens in contrast are often
installed directly in the way of passers-by, such that they
naturally appear in their field of view. Müller et al. [23]
show that what people expect on public displays depends
on the immediate surroundings, e.g. the apparent owner of a
shop window where the display is installed. As columns are
usually freestanding, expectations may differ. Brignull and
Rogers [5] and Peltonen et al. [24] observed that users were
waiting for their turn to access the display. For cylindrical
displays, there is no central position where the user could
‘own’ the display, but all positions around the display are
equal. Therefore, turn taking could be very different.
Influence of the Display Frame on Audience Behavior
A major difference between planar and non-planar screens
may result from their frame. In this context Manovich [20]
presents a theory of the imprisonment of the viewer’s body
by the screen apparatus on one hand and the requirements
of the image perspective on the other hand. In cinema, e.g.,
the body of the viewer is confined to a seat and the head is
aligned to forward view hence providing the best view-
point. The same is true in classical arts (e.g., a painting in a
museum) where the viewers seem to position themselves
centrally at some distance in front of the screen. We show
that this effect does not appear in front of non-planar
screens. Another interesting conception of Manovich is the
description of the screen’s frame as a clearly defined rec-
tangle, constituting a “viewing regime”. Anything outside
the frame can be ignored by the viewer, while immersing
himself into the content inside the frame. Pinhanez and
Podlaseck [25] discussed advantages and disadvantages of
frameless displays, also claiming the significance of the
frame to serve as a reference for the viewer to orient inside
the scene and position himself accordingly. Cylindrical dis-
plays introduce a new category in between framed and fra-
meless displays. It this case of semi-framed displays, a
frame is provided on the top and bottom, but not left or
right.
PROTOTYPE OF A CYLINDRICAL DISPLAY
For our research we developed a prototype of an interactive
cylindrical display. Though our prototype is projector-based
we envision that future versions will be based on bendable
display foil. In the following, the hardware and software
setup of this display is being described.
Hardware
The prototype of the cylindrical display consists of a cluster
of 8 standard projectors, 4 foil mirrors, 10 PCs, and a rear-
projection screen. It has a height of 2.2 meters and a diame-
ter of 1.5 meters. The 4:1 projection screen is 1.1 meters
high, has a diameter of 1.3 meters and a resolution of 2048
to 512 pixels. Each cluster element projects onto a mirror
reflecting the projection onto about one quarter of the
screen. For a viewer independent image blending we use a
special rear projection screen with a low gain factor. To de-
Figure 2: Prototype of a cylindrical display: (1) Camera
sensor. (2) Fisheye lens. (3) Acrylic rear-projection screen.
(4) 4 x Foil mirror. (5) 8 x Standard projector of 1024x768
pixels.
Figure 3: Cylindrical display with flower content for pass-
ing-by interaction as used in our user study.
tect position and movement of the passers-by around the
column, a sensor interface is installed above the column
consisting of a high-resolution camera and a 185° fisheye
lens. The hardware setup is depicted in Figure 2.
Software
The raw projection on the cylindrical screen is heavily dis-
torted due to the curvature of the screen and the projection
angle. To correct these distortions we use the calibration
technology described in [3]. This software is also used to
resolve the blending function for the real time correction
between overlapping regions of adjacent projectors. The
visualization software for the displayed contents on the col-
umn screen is based on a distributed rendering system. To
enable user interaction, we implemented a motion tracking
software, using OpenCV. The motion tracker uses frame
differencing to detect motion, and calculates the angle,
speed, and pixel distance of moving blobs from the column.
The Kalman filter is used to smooth these trajectories.
We also developed a VRML-based application framework
that enables us to display the same kinds of interactive ap-
plications within the coordinate systems of flat as well as
round screen shapes. For the interactive digital column dif-
ferent sample applications were designed (see Figure 4):
1. Reactive animated typography. In this application
words appear on the column as the user walks around
it. While moving around the column it’s in principle
possible to explore content from left to right as well as
from right to left. To make the text readable while
passing clockwise as well as counterclockwise, both
the horizontal and vertical dimensions are used. In or-
der that words don’t collide when taking a shift in di-
rection, the text is presented on a visual step-curve. To
ensure viewers do not lose track of the text flow, the
next appearing word is always announced by an ani-
mated dot.
2. Direction of the viewer. In this application items are
flying “towards” the user in z dimension, following
the user by adjusting their horizontal direction. For
example, the crown of a soft drink bottle seems to fly
from the inside to the outside of the column and target
the viewer, who cannot evade the item as he is tracked
by the camera sensor. Such content, where the viewer
cannot disregard the information, can be used for pre-
senting urgent or provoking messages in social con-
science campaigns for example.
3. Tales around the column. In this application, an end-
less picture story of drinking up a soft drink bottle is
told around the column. With the help of the camera
sensor the application presents the first picture on the
side where the viewer is approaching, and animates
the viewer to move further around the column and
proceed in the story by providing information stepwise
only. In order to see the end of the tale, the user has to
circle the column.
4. Movements of the viewer. In this application, the user
paints a pattern of flowers or bubbles on the column
by any kind of body movement. If the user stops the
flowers slowly fade away, and the individual pattern
that has been painted can’t be seen any more. Yet,
when moving on the flowers reappear and follow the
user as he moves along the surface, so that he is en-
couraged to proceed moving.
5. Reactive ambient column. In this application, each
person approaching the column is represented by “be-
ing served” a soft drink bottle, which appears in the
direction the person is standing. The aim is to give in-
formation about the public gathering and creating a
social atmosphere. There have been reactive media fa-
çades in the past that present ambient information
about the weather or about what is happening inside a
building (e.g., a train station). The round nature of the
column also allows mirroring objects in the outdoor
space around the column, while a flat screen would
only be able to cover one spatial direction.
Figure 4: Developed applications for the cylindrical screen:
(1) Reactive animated typography. (2) Direction of the
viewer. (3) Tales around the column. (4) Movements of the
viewer. (5) Reactive ambient column. (6) Interactive multi-
player game (detailed descriptions below).
6. Interactive multiplayer game. This is a game where
players are represented by individually colored bub-
bles on the screen appearing when moving fast, and
have the task to burst all their opponents’ bubbles.
This game is making use of the fact that multiple users
can interact with the round body of the column at the
same time. In general all kinds of games are imagi-
nable where teammates or opponents can chase each
other around the column or hide behind it, like cops
and robbers, hide and seek, etc.
Of these available applications, we chose application 4 for
our study. This was because the application works for a
single user and we were interested in user movement. The
chosen application draws a flower pattern on the screen
when the user moves, and he can influence the horizontal
and vertical position of the flowers by walking left or right
or by waving his hand. He can also influence the size of the
flowers by moving faster or slower (see Figure 3).
HYPOTHESES
As no studies regarding user behavior towards cylindrical
displays exist, first of all it is important to understand how
people move and behave around them. This knowledge can
then be used as a basis to develop applications that exploit
the properties of the new format and investigate more ela-
borate topics such as multi-user interaction. As there are
many (real-world) situations where only a single user is in-
teracting, we decided to concentrate on single-user interac-
tion only, leaving multi-user interaction for future work.
Based on informal observations of colleagues and visitors
we posed three general hypotheses that characterize behav-
ior around cylindrical displays. First, we assumed that users
walk more when interacting with cylindrical than with flat
displays. If true, this is an important property, as many flat
displays are designed for people standing in front of them.
Designs that work for people standing, with rather high
complexity and small fonts, may not work for people walk-
ing. Second, we assumed that while users seem to have
their shoulders parallel to flat displays, they would have
their shoulders in a certain angle to columns. This would be
an important property for gesture-based interaction, since
while users can use both arms equally for flat displays, one
arm would be turned away from a column, making symmet-
ric gestures difficult. Further, it would be difficult to move
any arm against the direction the user is facing, so a whole
different gesture set would need to be designed. Finally, we
hypothesized that due to the more active engagement users
would spend more time interacting with columns.
Hypothesis 1: Users walk more when interacting with
the cylindrical display.
1a: Users walk longer distances when interacting with the
cylindrical display.
1b: Users spend more time walking when interacting with
the cylindrical display.
1c: The position of users has a higher variance when in-
teracting with the cylindrical display.
Hypothesis 2: Users position themselves with shoulders
parallel to the flat display but not to the cylindrical dis-
play.
2a: The users’ shoulder position is parallel to the display
less often when interacting with the cylindrical display
while walking.
2b: The users’ shoulder position is parallel to the display
less often when interacting with the cylindrical display
while standing.
Hypothesis 3: Users spend more time overall interacting
with the cylindrical display.
In addition, we had several hypotheses for the viewing be-
havior of participants. We hypothesized that participants
would look more often at the cylindrical display, but for
shorter bursts. We also hypothesized that participants would
look at the left half of the column when walking clockwise
and the right half when walking counterclockwise. While
these hypotheses were formed before the design of the user
study, we additionally conducted a post-hoc analysis of the
data to explore further observations we made.
USER STUDY DESIGN
In order to test these hypotheses we conducted a user study
comparing single users’ behavior in front of interactive flat
and cylindrical displays. The study was conducted at our
lab over the course of two days. In the following chapter we
describe the design and setup of our user study and report
on the recruiting process as well as the study procedure.
We opted for a lab study due to the following reasons. (1)
For the anticipated measurements, a highly controllable en-
vironment was required allowing for statistical data analy-
sis. This would have been difficult to achieve in public due
to a high amount of external influence and fragile, technical
equipment. (2) To assess the users’ behavior we used cam-
eras during the study. This would have been a major issue
in public due to privacy reasons. To create an authentic
scenario we created a situation where participants were free
to visit different rooms containing various exhibits. Hence
we created a situation where (1) people behaved in a semi-
natural way, (2) people were not aware what we were
measuring in order to avoid influencing their behavior, and
(3) we created a controllable, yet still realistic scenario.
Setup
For the study we prepared 4 rooms at our lab each of which
contained a prototype. Two of the rooms contained “fake”
prototypes, which were functional (one was an interactive
flat screen with content that reacted to the viewers’ head
movements and facial expressions, the other one was a non-
interactive dome projection and showed a movie) – how-
ever their only purpose was to create a more realistic situa-
tion and to distract from the displays under investigation.
