[Next][Up][Previous]
Next:Real
Time detection Up:Virtual
Kathakali: Gesture-Driven Previous:Introduction
Outline
The system can be broken up into three modules:
-
Real-time Detection of arm movements of the user (Section 3).
-
Modelling of the Kathakali dancer (Section 4).
-
Reproduction of the pose in the Kathakali dancer's model (Section 5).
Unlike other models that use thermal imaging to obtain the user's silhouette
[7], the Virtual Kathakali system
uses a visible-light monochrome camera against a black background. The
user's silhouette is obtained by dynamically binarizing the images and
the 3-D positions of the user's shoulder, elbow and the wrists are obtained
in real time from the image coordinates. This compact data is then transmitted
to a local or off-site virtualization system in real time. Also, by using
skin-tone colour/greyscale information, it is possible to identify occlusion,
which is not possible to do in thermal imaging systems. The overall cost
of this system is likely to be several times less than that of other comparable
systems used in Virtual Metamorphosis systems.
The next phase is to create an articulated 3D model that will follow
the user's poses and reflect the traditional costumes of a classical Indian
dance form such as the Kathakali. The 3D model needs to have appropriate
motion constraints at the joints and suitable dress/headgear/texture. The
3D arm pose sequence is now communicated to the graphics model, which recreates
it as an animated graphics display. In this process, the very low-bandwidth
joint angle data can be used to animate the 3D Dancer model.