### Summary
This reading describes the problems of computer puppetry, the mapping of a performer to an animated character which may have a different geometry.  It introduces methods for coping with these difficulties in real-time.
### Problem
Basically, when transferring captured performance data to a character with different dimensions from the performer, only the joint angles or end-effector positions can be retained, but not both.  The problem this reading seeks to address is how to determine which of these are more important at any particular point in time and satisfy the correct one accordingly.
### Methods Used
This system defines importance rules to determine what parameters to preserve:
* The root position is not considered important
* An end-effector interacting with the world is considered important
* An end-effector which is probably about to interact with the world is considered important
* If an end-effector is moving away from an object, it's not considered as important
* An end-effector not near any objects is not considered important

To evaluate these rules in real-time (and to reduce the effect of noise in the captured motion data), it uses a Kalman filter to predict future poses based on past poses.
### Key Ideas
Kalman filtering of motion capture data for real-time feedback
An importance system for determining which captured parameters need to be preserved