el 
Of 
JT 
nd 
for 
in, 
for 
3al 
va, 
on 
SS, 
hi 
ite 
nal 
  
ISPRS Commission III, Vol.34, Part 3B „Photogrammetric Computer Vision“, Graz, 2002 
  
“THE RIGHT MOVE” — A CONCEPT FOR A VIDEO-BASED CHOREOGRAPHY TOOL 
S. Zillner?, M. Gelautz^, M. Kallinger" 
* Institute for Computer Science and Business Informatics, University of Vienna, 
Liebiggasse 4, A-1010 Vienna, Austria - sonja.zillner(gunivie.ac.at 
? Institute for Software Technology and Interactive Systems, Vienna University of Technology, 
Favoritenstrasse 9-11/188/2, A-1040 Vienna, Austria - gelautz@ims.tuwien.ac.at 
KEY WORDS: Video, Multimedia, Biometrics, Human Motion, Choreography, Dance 
ABSTRACT: 
In the context of a proposed project for a video-based choreography tool, we carry out video acquisition experiments with a Sony 
DCR-VX2000 camcorder that operates at 25 frames/sec and a high-speed HCC1000 video camera with adjustable frame rates. We 
perform tests with frame rates up to 462 frames/sec and observe that the performance of the Kanade-Lucas-Tomasi tracking 
algorithm improves significantly as the frame rate increases. Furthermore, we demonstrate the application of an available dance 
notation, the Labanotation, to the video-captured ballet scene. 
1. INTRODUCTION AND MOTIVATION 
Capturing human shape and motion from image sequences 
constitutes an important topic in current computer vision 
research. The recognition, analysis, and realistic modelling of 
people from videos is required in a wide range of applications, 
including man-machine interfaces, the creation of human 
avatars, medical and sports analysis, or safety applications. 
The proposed project focuses on complex movements of the 
whole body as presented in the field of classical dance (ballet). 
The subtleties of the human dance movement pose special 
challenges during the motion capture and analysis procedure 
which are not encountered in studies that concentrate on, e.g., 
human walking. The development of a tool that supports the 
(semi-)automatic recording and annotation of dance sequences 
would be of great interest to choreographers and dance 
companies in order to achieve a reduction in time and cost of 
rehearsals. This is particularly important if an existing repertoire 
Is going to be put on stage again with a different choreographer 
and/or new dancers. Beyond that, an archive of recorded dance 
scenes along with extracted 3D position and motion data can be 
utilized later by the choreographer as a computer-based creative 
tool as well as for animation purposes. 
Video-based motion capture for dance applications faces a 
series of challenging questions such as 
- What are the characteristic features of classical dance 
movements and which of them can be extracted from the 
video data? 
- What would a suitable video acquisition set-up look 
like (eg, number and type of cameras, spatial 
arrangement, use of markers, etc.)? 
: Is it possible to classify automatically the recorded 
video scenes (e.g., to recognize all pirouettes?) 
In order to describe the characteristics of a ballet sequence, we 
will make use of available movement/dance notations (e.g., 
Labanotation), whose design goal was to provide a concise 
description of the human body movement. These dance 
notations may, for example, be exploited in order to determine 
good marker positions on the human body. Section 3 describes 
the Labanotation (Hutchinson 1991) - named after the Austrian 
dancer and choreographer Rudolph Laban (1879-1958) - which 
is nowadays routinely utilized by choreographers and dance 
notators in order to annotate human body movement. In section 
4, we report tests with different video cameras and examine the 
effect of varying frame rates on the performance of an optical 
tracking algorithm. 
2. PREVIOUS WORK 
A survey of different computer-vision techniques for human 
motion analysis can be found in, e.g., (Gavrila 1999). Most 
related research has been performed in the context of modelling 
human movement for realistic computer animation (Liu et al. 
1999, Plaenkers and Fua 2001). Only few publications have 
addressed the field of dance. For example, (Suzuki et al. 2000) 
aim to develop an interactive dance system that takes into 
account the emotional state of a reference dancer by analyzing 
his/her body motion. (Abouafa/b 1999) deals with motion 
capture techniques in order to set up a dance performance in 
which real artists interact with virtual dancers. (Moltenbrey 
2001) describes the use of an optical motion capture system to 
record the movements of famous pantomime Marcel Marceau in 
order to preserve the artist's legacy for future generations. 
A relatively new area of application is the development of 
human-computer interfaces which take into account the human 
body language. A related experiment is reported by (Nakata et 
al. 1998), who propose to express a robot's emotion by its 
posture and dance. The work is based on Laban's theory on 
motion analysis and description. A system for generating 
Labanotation from motion captured data is proposed by 
(Matsumoto et al. 2001). 
B - 313