Imitating Complex Realistic Gestures

Human adults are amazingly skilled at observing and then imitating one another's actions. In fact many important skills are acquired and refined by means of imitative learning. As part of the NSF-sponsored Center for Excellence in Learning, Education, Science and Technology (CELEST) members of the lab at Brandeis are taking a variety of approaches to unraveling the important and fascinating mysteries of imitation.

Some of the lab's imitation research exploits the unique theoretical value of abstract two-dimensional models, examples of which can be seen here .

Recently, we started to study imitation of complex gestures made by realistic, computer-generated human models. Some of the gestures are modeled after gestures from American Sign Language (ASL). One example can be seen at the left. Watch carefully as Masha, the model shown here, executes a gesture. Then see how well you can imitate her actions. A warning: this QuickTime movie is about 200 KB, so it may not load instantly, but it's worth the wait. After the movie's loaded, click on the arrow at the left side of the controller to set Masha into action. After you've seen Masha's gesture once, try to imitate what she did. Then have her repeat the gesture, and compare it to your version. Each time you watch, your imitation will probably improve. Note what aspects of the gesture you pick up on successive viewings. If your imitation is perfect the very first time, you can be really proud --that's rare.

The lab has two complementary motion-capture systems for recording and analyzing the movements people make as they struggle to imitate what they've seen. This tryptich shows part of one system for recording a person's arm, hand and fingers movements.