Advances in Vision-Based Human Body Modeling - 3D Modeling and Animation: Synthesis and Analysis Techniques for the Human Body - page 18

walking and running speeds and direction of motion. One of the constraints is that
the motion must be front-parallel. Gavrila & Philomin (1999) present a shape-
based object detection system, which can also be included into the surveillance
category. The system detects and distinguishes, in real-time, pedestrians from a
moving vehicle. It is based on a template-matching approach. Some of the
system's limitations are related to the segmentation algorithm or the position of
pedestrians (the system cannot work with pedestrians very close to the camera).
Recently Yoo, Nixon & Harris (2002) have presented a new method for
extracting human gait signatures by studying kinematics features. Kinematics
features include linear and angular position of body articulations, as well as their
displacements and time derivatives (linear and angular velocities and accelera-
tions). One of the most distinctive characteristics of the human gait is the fact
that it is individualistic. It can be used in vision surveillance systems, allowing the
identification of a human by means of its gait motion.
User Interface
User interface is another application domain that takes advantage of 3D human
body modeling. Wingbermuehle, Weik & Kopernik (1997) present an approach
to generate highly realistic 3D models of participants for distributed 3D
videoconferencing systems. Using 3D data obtained by means of stereoscopy,
the size and shape of each real person is recovered and represented through a
triangular mesh. In addition, texture extracted from the real images is mapped
to the 3D models leading to a natural impression. Together with a flexible
triangular mesh, a skeleton structure of the human model is build. The latter is
used to preserve the anthropomorphic constraint. Cohen, Medioni & Gu (2001)
present another real-time 3D human body reconstruction for vision-based
perceptual user interface. The proposed system uses multiple silhouettes ex-
tracted automatically from a synchronized multi-camera system. Silhouettes of
the detected regions are extracted and registered, allowing a 3D reconstruction
of the human body using generalized cylinders. An articulated body model
(defined by 32 DOF) is fitted to the 3D data and tracked over time using a particle
filtering method. Later on, Cohen & Lee (2002) presented an extension of this
work that consists of an appearance-based learning formalism for classifying
and identifying human postures.
Davis & Bobick (1998a) present a novel approach for extracting the silhouette
of a participant within an interactive environment. This technique has been used
in Davis & Bobick (1998b) for implementing a virtual Personal Aerobics Trainer
(PAT). A computer vision system is responsible for extracting the human body
movements and reporting them to a virtual instructor. With this information, the