'Facing' the Future
I'm Bob Karson with the discovery files -- new advances in science and engineering from the National Science Foundation.
Meet Einstein, a robotic head at UC San Diego that's a dead ringer for the father of the theory of relativity. This robotic Einstein is learning facial expressions -- machine learning. Instead of being programmed to make only certain specific facial expressions, it can create new combinations.
(SOUND EFFECT: baby babbling) Developmental psychologists believe that infants learn speech and control of body movements through systematic exploration. To learn to speak, we first babble -- to learn motion, we try all kinds of clumsy movements.
The team applied this principle, called "body babbling" to robotics. The 30 facial 'muscles' that make Einstein smile, frown, or look surprised, are controlled by strings attached to tiny servo motors. They first had Einstein twist and contort his face every which way while 'watching' himself in a mirror, and relaying that information to facial recognition software. This maps the changes that are possible, and 'teaches' the robot the relationships between motor movements and facial expressions.
It could then create new expressions it never encountered. Like moving the inner eyebrows together and slightly closing the eyelids.
Aimed at making robotic facial expressions more realistic, the research may also tell us more about human facial movements. He's looking at me funny.
"The Discovery Files" covers projects funded by the government's National Science Foundation. Federally sponsored research -- brought to you, by you! Learn more at nsf.gov or on our podcast.