Our research into vision-based user interfaces is
trying to answer questions like:
-
Can a camera be used as a user interface device,
much as we use a keyboard or mouse?
-
Can computer vision
infer human actions with sufficient accuracy and reliability?
-
Can we map those actions to computer
control tasks in a natural way?
-
Is this a useful way to interact with
a computer?
We have developed a number of fundamental interaction types,
such as touching a target or pointing your face, that we can recognize with
good reliability. These actions can then be mapped to user interface tasks such
as triggering events or moving a pointer. We are also exploring the related areas
of face recognition and visual assists for voice recognition.
In the near
term we are developing stand-alone tools that work with current user interfaces,
many targeted at the special needs community.
In the long term we are working toward an integrated "direct
sensing" or "deviceless" user interface composed of vision-based
modalities plus voice recognition and perhaps others.
|
