Determining the driver
state is important for detecting driver distraction or
inattention. Facial action codes, eye blinks, and head movements
are all cues for driver state. By analyzing these cues we can make
a safe driver experience. Additional research has been performed
on using these cues to add context to remote communication. An
example of this would be transmitting the drivers expression over a cell
phone to a remote user.
Robust expression estimation can be performed using thin-plate
splines. This allow the non-rigid movements from expressions to be
separated from the rigid body head movements efficiently. Here is
an example of the thin-plate spline transformation (click on the image
for a video).
Here are some of the results from the expression estimation system
(click on the image for a video).