Each phase involved a specific set of challenges. Phase 1 (segmentation
of existing animation into action units and labeling)
proved that identifying meaningful actional units for reuse in an animation
clip is a reasonable task. Given an animation composed of 40
seconds (1000 frames), the animator identified 43 actions (e.g., writing,
grabbing, etc.) and 27 poses (e.g., surprise takes, eye blinks, etc.),
most of which involved only the facial expressions.