User Awareness for Collaborative Multi-touch Interaction

Markus Schlattmann

, Yuelong Yu

, Nils Gruendl

, Manfred Bogen

, Alexander Kulik

David d'Angelo

, Bernd Froehlich

and Reinhard Klein

AGT Group (R&D) GmbH, Darmstadt, Germany

Fraunhofer IAIS, Sankt Augustin, Germany

Bauhaus-Universität, Weimar, Germany

University of Bonn, Bonn, Germany

Keywords: Multi-touch, User Aware, Applications, Interaction Metaphors.

Abstract: Multi-touch enables direct manipulation of graphical computer interfaces. This intuitive interaction

paradigm rapidly became popular, particularly in the domain of individual mobile devices. Collaborative

work on large scale multi-touch devices, instead, suffers from mutual interferences if many simultaneous

touch events cannot be attributed to individual users. On the example of a large, adjustable, high-resolution

(4K) multi-touch device, we describe a lightweight and robust method to solve this issue. An additional

depth camera above the tabletop device tracks the users around the table and their respective hands. This

environment tracking and the multi-touch sensor are automatically calibrated to the same coordinate system.

We discuss relevant implementation details to help practitioners building similar systems. Besides improved

multi-user coordination we reveal general usability benefits including the reduction of false positives.

Exploiting the developed system, we implemented several novel test applications to analyze the capabilities

of such a system with regard to different interaction metaphors. Finally, we combined several of the

analyzed metaphors to a novel application serving as an intuitive multi-touch application and environment

for seismic interpretation.

1 INTRODUCTION

Many professional computer applications could

benefit from immediate collaboration of colleagues

and require technologies that support this. In the

particular realm of seismic data interpretation in the

oil and gas industry, we observed several issues

originating from isolated desktop workplaces

including the loss of information and the

misinterpretation of results.

Tabletop computers with multi-touch (MT) input

offer a promising alternative. They can be easily

operated by non-computer scientists and facilitate

smooth communication and cooperation among

different experts.

Unfortunately, most existing multi-touch systems

suffer from missing user awareness. If a multi-touch

system detects two touch points on the screen, it

generally cannot distinguish whether the touch

points belong to one hand, two hands or even

different users. Therefore, multiple users and

multiple hands can only work in the same context,

which often results in interference.

To solve these problems and allow for more

natural collaboration, some previous research

systems already included additional sensor

technologies. However, these systems all suffered

from severe limitations, restricting the quality of the

tabletop display or its surrounding either, or even the

movements/locations of the users themselves. Using

a depth camera (e.g. Microsoft Kinect), instead,

provides additional information that enables reliable

and robust user tracking. This in turn enables the

robust association of detected touch-points to

individual users and their hands, which provides

many new possibilities to improve the

expressiveness of multi-touch gestures and realize

software-supported multi-user multi-touch input

coordination. In particular we identify the following

applications:

 Individual users can associate different tools to

their input; thus different functions may even be

operated simultaneously by cooperating users.

359

Schlattmann M., Yu Y., Gruendl N., Bogen M., Kulik A., d’Angelo D., Froehlich B. and Klein R..

User Awareness for Collaborative Multi-touch Interaction.

DOI: 10.5220/0004210603590366

In Proceedings of the International Conference on Computer Graphics Theory and Applications and International Conference on Information

Visualization Theory and Applications (GRAPP-2013), pages 359-366

ISBN: 978-989-8565-46-4

 2013 SCITEPRESS (Science and Technology Publications, Lda.)

 Software-controlled access management eliminates

involuntary interference (e.g. during manipulation

an object access is blocked for other users)

 Automatic partitioning of the screen and input

space with respect to user positions.

The main objective of our work was the

development of a functional tabletop prototype that

provides all the features and the quality required for

co-located collaboration on the interpretation of

seismic data. For the moment we were focusing only

on the acceptance of the system by expert users. In

future work we are aiming to gain further insights

into the usability of the system and its integration in

the professional workflow.

2 RELATED WORK

Many researchers implemented multi-user

coordination policies using the commercially

available DiamondTouch system (Dietz and Leigh,

2001); (Morris et al., 2004); (Morris et al., 2006);

(Morris et al., 2004). Unfortunately, the system

limits the choice of display component to front

projection. It is furthermore limited to four users that

must maintain physical contact to the corresponding

receiver unit while avoiding to touch each other.

Marquard et al., (2010) used a glove that tagged

features of the hand with unambiguous optical

markers for robust hand and finger identification.

Roth et al. propose attaching small IR-emitting

devices to the users' hands for cryptographically

sound user identification (Roth et al., 2010). This is

obviously the most secure implementation of user

tracking for tabletop interaction, but requiring the

users to wear an electronic device which breaks with

the walk-up and use paradigm of most tabletop

applications.

Dang et al., instead, proposed a system that does

not require the user to wear a glove. They suggested

a heuristic to identify the hands belonging to a touch

point based on the orientation of the tracked ellipse

(Dang et al., 2009). Besides the limitation that users

must always touch the surface with the finger pad

instead of the tip, the association is immediately lost

once the fingers are released from the screen. More

recently, Ewerling et al. proposed multi-touch

sensing based on the maximally stable extremal

regions algorithm that implicitly organizes touch

points in hierarchies corresponding to individual

hands (Ewerling et al., 2012). However, this

approach only works with additional depth

information above the tabletop and cannot

distinguish individual users.

Towards the association of touch points to users,

Walther-Franks et al. proposed to use additional

proximity sensors in the frame of the tabletop device

(Walther-Franks et al., 2008). The system provides

rough user tracking for many purposes, but due to

the dislocation of the touch sensing on the screen

surface and the user tracking around its housing, a

robust correlation of touch-points with a user's hand

cannot be ensured. Touch points detected in close

proximity to the user's body position, tracked at the

edge of the tabletop device, may also belong to

somebody else reaching into her proximity.

Recently, Annett et al. presented an improved

version of such a system, using a much larger

number of proximity sensors to derive a higher

tracking density (Annett et al., 2011). In particular,

their system incorporates sensors in upward

orientation, tracking the user’s arm above the

display frame. This adaptation enables a more

accurate association of tracked touch points to

individual users. While this system is a significant

improvement, the general limitations related to the

missing overlap of both tracking systems remain. In

a similar spirit, Richter et al. (Richter et al., 2012)

suggested to capture the users’ shoes for

identification. The association with touch points

cannot be realized robustly, but individual settings

can automatically be applied based on the coming

and going of users - if they keep wearing the shoes

that are linked to them in the database.

Dohse et al. used an additional camera mounted

above the tabletop display for the association of

touch-points with users. The hands are tracked

above the display using color segmentation or

shadow tracking respectively (Dohse et al., 2008).

The authors suggest cancelling the light from the

screen with polarization filters to avoid interference

with the color of displayed items. As an alternative,

they propose tracking the dark silhouettes of the

hands above the illuminated screen.

Another approach by Andy Wilson suggests

using a depth sensing camera attached to the ceiling

as an all-in-one sensor both for touch detection with

defined interaction surfaces and context awareness

(Wilson, 2010). An advanced follow up on this

approach has been presented by Klompmaker et al.,

(2012). Their framework provides the means for

tracking tangible input devices and the users’ hands

for gesture recognition and touch input on arbitrary

surfaces. In both cases, the same sensor that is

exploited for touch sensing covers the whole

surrounding area in depth, touch data can directly be

associated with tracked users. As a downside of this

concept, the relative low resolutions of the depth

GRAPP2013-InternationalConferenceonComputerGraphicsTheoryandApplications

360

cameras impede accurate multi-touch input.

Martinez et al., (2011) consequently proposed to

combine the context tracking of the depth camera

with accurate touch sensing at the surface of the

tabletop device. Unfortunately, they did not provide

much detail about their implementation.

Jung et al. (Jung et al., 2011) proposed the

combination of the Kinect depth sensor for body

tracking and RFID for the automatic authentication

of users on a multi-touch tabletop device. They

realized a prototype demonstrating the benefits of

the concept for applications with critical security

issues. The overall usability of the system and the

technical details necessary to achieve this were not

discussed in the paper.

3 MULTI-TOUCH CONTEXT

TRACKING

In this section, the method for computing the

association of touch-points with respective users and

hands is described.

3.1 Hardware Setup

Our high-resolution multi-touch table (3840x2160

pixels) measures 56" display diagonal. With the

Kinect device mounted about 3 meters above the

floor (see Figure 1) we can capture the entire screen

area and about 50 cm of its surrounding in each

direction.

Figure 1: The arrangement of MT-table (Bottom) with

Microsoft Kinect (Top).

A standard desktop PC is used to drive the

application on the multi-touch table. A second

machine runs the environment tracking. Both

workstations share the recorded user input data via

network using the TUIO protocol (Kaltenbrunner et

al., 2005).

3.2 Automatic Calibration and

Calibration Detection

The internal calibration of the different sensors

embedded in the Kinect (color, infrared, depth) must

be performed only once as their relation does not

change over time. We use the method published by

Nicolas Burrus for this purpose (Burrus, 2011).

The extrinsic parameters defining the relation

between the display area and one of the Kinect

sensors, instead, must frequently be re-calibrated.

These parameters have to be re-computed every time

the display or the Kinect has been moved. Due to the

adaptability of our assembly this can happen quite

often. For ergonomic usage in different situations

our touch table allows to adjust the height (75 - 125

cm) and the orientation (0 - 70 degrees) of the

interactive display. Even the pivot of the display can

be changed between landscape and portrait

orientation.

We propose a fully automatic calibration routine

without the necessity of any additional calibration

object such as a printed chessboard pattern. The

screen itself is employed to display a calibration

pattern (chessboard) that can be recorded by the

color sensor of the Kinect and processed for

calibration with the method available in OpenCV

(Bradski, 2000). Note that the calibration pattern on

the display is not visible for the infrared or depth

sensor of the Kinect. Additionally, the need for re-

calibration induced by height or inclination changes

is detected automatically using motion sensors

attached to the multi-touch table.

3.3 Combined Segmentation

The sensors of the Microsoft Kinect enable different

ways of segmenting the foreground (user body parts)

from the background (e.g. floor and display).

However, every sensor and according method has its

benefits and drawbacks. We achieved the best

results with a combination of the IR-intensity image

and the depth information. Depth segmentation

facilitates the background subtraction in the

uncontrolled surrounding of the tabletop device.

However, the imprecision and quantization of depth

values obtained by the Kinect impede precise depth

segmentation close to the display surface.

UserAwarenessforCollaborativeMulti-touchInteraction

361

Furthermore, the effective image resolution of the

depth image is low (approximately 320x240),

wherefore small body parts (e.g. fingers or small

hands) tend to vanish in the segmentation (see Figure

2(Left)).

Figure 2: A combination of depth segmentation (left

image) and IR-intensity segmentation (middle image)

leads to better segmentation results (right image) and

solves the respective problems (see regions in green

circles). The red circles indicate the problematic regions in

depth-based segmentation (left) and IR-based

segmentation (middle), respectively.

The background subtraction performed on

infrared (IR) intensity image from the Kinect

provides a higher resolution (640x480) and also

works close to the display surface as the display

does not emit infrared light. Using a logical or-

operation, we combine the depth-based

segmentation of the entire image with the infrared

segmentation, but only inside the image region

corresponding to the display surface. The result is

shown in

Figure 2.

3.4 User Separation and Tracking

After the segmentation, only noise and those regions

belonging to the users remain in the image. To filter

high frequency noise we first perform an erosion

step followed by a dilatation step. Then, we search

for the largest connected components. A simple

threshold on the minimum component size filters

further artifacts of noise. The remaining connected

components (CC) can be assumed to correspond to

individual users. They are assigned a user ID and

tracked over time (Figure 3).

A drawback of this approach is that as soon as

two users touch or occlude each other, their separate

CCs will merge. Thus, we apply a second processing

step to separate the user regions also if a CC

comprises multiple users. We exploit the depth

information provided by the Kinect camera to

identify the upper body of each user.

Our algorithm searches for height peaks within

each connected component that are at least 40 cm

above the known height of the tabletop surface. If

only one of such peaks exists, the entire component

is interpreted as a single user region. Otherwise, the

region is separated into individual areas surrounding

each peak based on frame coherence (i.e. pixels of

the CC are divided based on the nearest pixel

assignment of the previous frame). Apparently, as

this solution is based on frame coherence, it only

works if at a certain time in the past the users did not

occlude each other. Furthermore, it sometimes leads

to erroneous separation in extreme cases, e.g. if the

occlusion continues for a long time during complex

movement of the users.

Figure 3: Example segmentation of four users.

3.5 Associating Touch-Points to Users

and Hands

For each touch-point the corresponding user and

hand have to be estimated. To this end, we project

the touch-points to the image containing the user

regions using the transformation matrix derived

from the calibration (see Figure 4).

Figure 4: Projection and assignment of the touch-points

(black dots on the left) to the image containing the user

regions (right). The blue rectangle indicates the area of the

display surface. The projected touch-points (colored dots

with black margin on the right) are assigned to the closest

user region.

If a touch-point is located inside a user region,

the corresponding user ID is transcribed directly.

Otherwise the closest user region is computed using

a breadth-first search. We further distinguish both

hands of a user based on the geodesic distances of

touch-points in the graph representing the user

component. Our procedure builds on the Dijkstra

algorithm to find the shortest path between the

touch-points assigned to one user. The procedure

stops either if all the other touch-points of the

respective user have been found or if the Dijkstra

radius exceeds a certain threshold D (30 cm).

GRAPP2013-InternationalConferenceonComputerGraphicsTheoryandApplications

362

The resulting clusters are interpreted as

individual hands. Unlike Euclidean distances,

geodesic distances support robust clustering even if

both hands are in close proximity. However, this

approach is still limited to cases where the hand

regions do not merge to one region in the camera

images. If the hands merge, we can only transcribe

the hand state of touch points from a previous state

(if existing).

3.6 Identification of Involuntary

Touches

If, for one respective user, more than two clusters

are found we ignore those with the lowest frame

coherence as the respective touch events most likely

occurred involuntarily. We also classify a touch-

point to be involuntary if the area of the touch

footprint on the multi-touch panel is larger than a

certain threshold (5cm²). This way, touches too large

to originate from a finger or soft-touch stylus are

ignored. This simple thresholding allows users to

rest their hands on the tabletop surface while

operating with their fingers or a stylus.

A one-time invalid touch-point stays invalid. We

also tried other mechanisms allowing touch-points to

become valid again such as (adaptive) hysteresis

thresholding. However, we observed this simple

mechanism to work best. We observed that

voluntary touches were hardly ever classified as

involuntary.

4 APPLICATIONS AND

FINDINGS

We first developed several test applications to

explore the novel possibilities of our user aware

tabletop system. These applications were informally,

but continuously tested by visitors of our lab in order

to identify which functionality best suites our

objective. Note that these demonstrators were

realized with an earlier multi-touch table prototype

based on a back-projection display. After these tests

we could identify the interaction techniques most

suitable for the seismic interpretation application.

We later implemented a novel system based on a

large high resolution LCD display. It was equipped

with a multi-touch sensor frame that provides high

precision and tracking robustness combined with

low latency. The final system including the user

tracking was set up with custom software for the

exploration and interpretation of seismic data.

4.1 Test Applications

Territoriality has been shown to be of particular

relevance for the coordination of multi-user

cooperation (Scott et al., 2004). The separation

between personal and group territories facilitate the

coordination of individual and cooperative subtasks.

In the context of a game based on the exchange

of virtual tokens we implemented a technique that

assigns a particular area of the shared tabletop

surface to each involved user (see

Figure 5). The

circular area is automatically placed in closest

possible position to the user it is assigned to and

serves as a storage area for managing the virtual

tokens. Other users could not access the collected

items.

Figure 5: The yellow circles with gray surrounding define

the users' personal storage space.

We observed that users immediately understood

the concept of the personal storage areas that were

following them. They appreciated that the areas

were always situated relative to their body, thus

facilitating access even while walking around the

table. We also frequently observed the annoyance of

users about the disappearing of their personal

storage area when they stepped away too far from

the table for a moment. We introduced some latency

to the automatic area management which alleviated

this issue a little. Only a robust user identification

could solve this problem (e.g. (Jung et al., 2011);

(Richter et al., 2012); (Roth et al., 2010)).

The orientation of GUI elements is a crucial

issue in the realm of tabletop interfaces. Users

generally surround the devices from all accessible

sides, while some elements might only be legible

from one direction. Shen et al. described a

metaphoric "magnet" feature allowing the

reorientation of all GUI-elements to improve

legibility. They observed user conflicts with this

global functionality (Shen et al., 2004). Turning all

items to face a particular user necessarily turns the

same items away from the others.

UserAwarenessforCollaborativeMulti-touchInteraction

363

We reasoned that users would generally not be

interested in orienting all available GUI elements

towards themselves, but only try to improve the

legibility of a few items they are focusing on. Thus

we implemented an automatism that turned a

selected item automatically towards the user who

selected it. This concept works well if users only

seek for their own reading comfort. In test

applications including a photo viewer and a

document reader, however, we observed that users

also want to show items to others. The automatic

alignment was interfering in these situations.

Therefore we adapted the algorithm such that items

still turned toward the user who selected them, but if

they are moved (e.g. in the direction of another user)

they turned toward the user being closest in the

movement direction. This combination was well

accepted by test users, but eventually we realized

that the items can just always be oriented in the

direction of a movement. This simplification works

similarly well and does not even require user

awareness.

Associating particular tools to the fingers of

individual hands is probably the most relevant

augmentation for multi-user multi-touch interaction.

In a drawing application we observed that users just

expected to keep a selected color for drawing until

they had chosen a new one (see Figure 6). In fact,

most of our visitors were puzzled when we

explained this to be a novel feature. They could not

imagine this to be any different. Consequently, they

were surprised that the chosen association was lost

when they stepped further away from the screen

such that they left the tracking area for a moment.

Figure 6: User awareness allows assigning individual tools

to each touch event. Here, each user draws with an

individually selected color.

Automatic input coordination can reduce

mutual interference of multi-user input enormously.

In the most basic configuration, GUI elements

cannot be acquired by users while they are already

manipulated by someone else. This ensures that each

user can complete a desired manipulation without

disturbance. We experimented with this feature in a

photo viewer application.

We found that more subtle adaptations to the

users' input are often preferable compared to simple

locking. In fact users often want to stop others from

moving an element. To this end they touch the

moving element in an attempt to stop it. Without any

context awareness such conflicting input from

multiple users generally results in scaling the

respective graphics element. We implemented a

coordination policy that only locks the type of

possible manipulations to those already operated by

the user who first acquired the element. Thus, if one

is moving an item, others may still interfere to hold

it, but this interference will not cause any other type

of transformation, e.g. scaling.

In combination with individual tool selection this

approach also allows multiple users to apply

different operations simultaneously on the same

element. One may for example continue to draw

lines on an object while it is moved around by

somebody else.

4.2 Seismic Interpretation Application

In the context of an industrial research project we

realized an application for the collaborative analysis

of seismic data on a multi-touch tabletop device. The

application builds on two of the above mentioned

techniques for user awareness. Furthermore, the

display quality was improved by using a high-

resolution LCD panel that offers multi-touch input.

Our earlier prototype did not satisfy the visual

quality required for the interpretation of the fine

grained data (see Figure 7).

Figure 7: Illustration of individual tool selection. The left

hand belongs to a user having enabled the line annotation

tool while the right hand belongs to a different user having

enabled the point annotation tool.

The application displays cutting planes derived

from a volumetric dataset. The collaborative task is

the interpretation of seismic features, searching for

places with a high probability of hydrocarbon

accumulation. The application primarily supports

navigation along a horizontal cutting plane as well

as selection and annotation of so called seismic

GRAPP2013-InternationalConferenceonComputerGraphicsTheoryandApplications

364

lines. In Figure 8 the collaborative work on such

seismic lines is shown for two users including the

depth image.

Figure 8: Left: Two users working on the same seismic

line. Right: According depth image.

From the above mentioned functionality we

implemented input coordination and individual tool

selection (see Figure 7). From our earlier

experiences with the test applications we expected

that technologically unaware users would not even

realize that they are tracked to ensure fluid multi-

user multi-touch interaction.

Experts from geology were using our context-

aware multi-touch system and we received very

positive feedback. Also the display size, its

impressive image quality with ultra-high resolution

and the accuracy of the multi-touch input were

highly appreciated. With the expectation of

becoming frequent users of the system our visitors

were delighted about the adaptation features of the

assembly, which easily allow changing the height

and inclination of the display. As expected, most of

these experts did not realize the implicit input

coordination to be an obvious feature. The system

just worked as expected.

5 CONCLUSIONS

We analyzed state-of-the-art methods to achieve

user awareness of multi-touch tabletop displays and

derived an improved method from our experiments.

Based on the sensor data from a depth camera we

achieved robust context tracking. We described an

implementation of this method including automatic

re-calibration, sensor-fused segmentation, separation

of users, robust hand identification based on

geodesic distances and a detection method for

involuntary touches. Based on the resulting user

awareness, we suggested interaction techniques for

fluent co-located collaboration. While the general

idea of context tracking with an overhead camera

has been proposed earlier, we contribute a detailed

description of a timely method that is robust and

easy to implement. Finally, we described a high-

fidelity system prototype and a collaborative

application for the exploration and interpretation of

seismic data.

We are looking forward to gain further insights

on the usability of the system in long-term studies

with expert users.

REFERENCES

Annett, M., Grossman, T., Wigdor, D., & Fitzmaurice, G.,

(2011). Medusa: a proximity-aware multi-touch

tabletop. Proc. UIST 2011 (pp. 337-346). New York,

NY, USA: ACM-Press.

Bradski, G., (2000). The OpenCV Library. Dr. Dobb's

Journal of Software Tools.

Burrus, N. (2011, June). Kinect RGB Demo v0.5.0.

Retrieved from http://nicolas.burrus.name/index.php/

Research/KinectRgbDemoV5

Dang, C. T., Straub, M., & André, E., (2009). Hand

distinction for multi-touch tabletop interaction. Proc.

ITS 2009 (pp. 101-108). New York, NY, USA: ACM-

Press.

Dietz, P., & Leigh, D. (2001). DiamondTouch: a multi-

user touch technology. Proc. UIST 2001 (pp. 219-

226). New York, NY, USA: ACM-Press.

Dohse, K. C., Dohse, T., Still, J. D., & Parkhurst, D. J.,

(2008). Enhancing Multi-user Interaction with Multi-

touch Tabletop Displays Using Hand Tracking. Proc.

Advances in Computer-Human Interaction 2008 (pp.

297-302). Washington, DC, USA: IEEE Computer

Society.

Ewerling, P., Kulik, A., & Froehlich, B., (2012). Finger

and hand detection for multi-touch interfaces based on

maximally stable extremal regions. Proc. ITS 2012

(pp. 173-182). New York, NY, USA: ACM-Press.

Jung, H., Nebe, K., Klompmaker, F., & Fischer, H.

(2011). Authentifizierte Eingaben auf Multitouch-

Tischen. Mensch & Computer 2011 (pp. 305-308).

München: Oldenbourg Wissenschaftsverlag GmbH.

Kaltenbrunner, M., Bovermann, T., Bencina, R., &

Costanza, E., (2005). TUIO - A Protocol for Table-

Top Tangible User Interfaces. Proc. of the 6th

International Workshop on Gesture in Human-

Computer Interaction and Simulation.

Klompmaker, F., Nebe, K., & Fast, A., (2012).

dSensingNI: a framework for advanced tangible

interaction using a depth camera. Proc. of ACM TEI

2012 (pp. 217-224). New York, NY, USA: ACM-

Press.

Marquardt, N., Kiemer, J., & Greenberg, S., (2010). What

caused that touch?: expressive interaction with a

surface through fiduciary-tagged gloves. Proc. ITS

2010 (pp. 139-142). New York, NY, USA: ACM-

Press.

Martínez, R., Collins, A., Kay, J., & Yacef, K., (2011).

Who did what? Who said that?: Collaid: an

environment for capturing traces of collaborative

learning at the tabletop. Proc. ITS 2011 (pp. 172-181).

UserAwarenessforCollaborativeMulti-touchInteraction

365

New York, NY, USA: ACM-Press.

Morris, M. R., Huang, A., Paepcke, A., & Winograd, T.,

(2006). Cooperative gestures: multi-user gestural

interactions for co-located groupware. Proc. CHI 2006

(pp. 1201-1210). New York, NY, USA: ACM-Press.

Morris, M. R., Ryall, K., Shen, C., Forlines, C., & Vernier,

F., (2004). Beyond "social protocols": multi-user

coordination policies for co-located groupware. Proc.

CSCW 2004 (pp. 262-265). New York, NY, USA:

ACM-Press.

Morris, R. M., Ryall, K., Shen, C., Forlines, C., & Vernier,

F., (2004). Release, relocate, reorient, resize: fluid

techniques for document sharing on multi-user

interactive tables. CHI Extended Abstracts 2004 (pp.

1441-1444). New York, NY, USA: ACM-Press.

Richter, S., Holz, C., & Baudisch, P. (2012). Bootstrapper:

recognizing tabletop users by their shoes. Proc. CHI

2012 (pp. 1249-1252). New York, NY, USA: ACM-

Press.

Roth, V., Schmidt, P., & Güldenring, B., (2010). The IR

ring: authenticating users' touches on a multi-touch

display. Proc. UIST 2010 (pp. 259-262). New York,

NY, USA: ACM-Press.

Scott, S. D., Sheelagh, M., Carpendale, T., & Inkpen, K.

M., (2004). Territoriality in collaborative tabletop

workspaces. Proc. CSCW 2004 (pp. 294-303). New

York, NY, USA: ACM-Press.

Shen, C., Vernier, F. D., Forlines, C., & Ringel, M.,

(2004). DiamondSpin: an extensible toolkit for

around-the-table interaction. Proc. CHI 2004 (pp. 167-

174). New York, NY, USA: ACM-Press.

Walther-Franks, B., Schwarten, L., Krause, M., Teichert,

J., & Herrlich, M., (2008). User Detection for a Multi-

touch Table via Proximity Sensors. Proceedings of the

IEEE Tabletops and Interactive Surfaces. IEEE

Computer Society.

Wilson, A. D., (2010). Using a depth camera as a touch

sensor. Proc. ITS 2010 (pp. 69-72). New York, NY,

USA: ACM-Press.

GRAPP2013-InternationalConferenceonComputerGraphicsTheoryandApplications

366