Goal
Fine designed robotic faces could make realistic facial expressions. However, manually tuning of the facial motor configurations costs hours of work even for experts. Is it possible for Robot to learn to do facial expression automatically by looking at himself?
In MPLab, we have developed computer vision based facial expressions recognition systems that enpower the robot to understand the definition of facial expressions. In particular, we use the definition of Action Units (AU) in the Facial Action Coding System (FACS). Then the ``Einstein'' is able to adjust his ``facial muscles'' to produce facial expressions by seeing his face in a mirror.
Learning Expressions Automatically
[HD Quicktime Movie]This text will be replaced
Manually Designed Expressions
Paper
- T.-F. Wu, Nicholas J. Butko, Paul Ruvulo, Marian S. Bartlett, Javier R. Movellan, “Learning to Make Facial Expressions,” International Conference on Developmental Learning (ICDL) 2009. [paper] [slides : PDF (no video) 15M, Quicktime 22M]