Authors:
Sindharta Tanuwijaya
and
Yoshio Ohno
Affiliation:
Keio University, Japan
Keyword(s):
Motion capture, Mocap, Camera.
Related
Ontology
Subjects/Areas/Topics:
Animation and Simulation
;
Animation from Motion Capture
;
Computer Vision, Visualization and Computer Graphics
;
Human Figure Animation
Abstract:
In recent years, the amount of mocap has accumulated due to its popularity in creating realistic human motions. However, such accumulation is yet to be accompanied by the development of a mocap search engine. In addition to the difficulty in processing mocap search, this phenomenon is also due to the problems in displaying mocap as search results, one of which is determining the camera position, orientation, and distance in displaying mocap. In this paper, we specify camera orientation and distance as constraints to determine camera positions by using available training data which are given as inputs into data mining techniques. In addition, we also discuss a method to select representative frames of mocap, thus allowing for the display of mocap search results as a list of sets of selected mocap frames. Finally, we employ a number of data mining techniques along with a simple method to determine the camera position which yields the widest projection area of a virtual face consisting
hands and feet joints into the camera plane, and compare the results to each other.
(More)