Cognitive Face Recognition
The first face recognition platform based on cognitive closed-loop
feedbacks is called FaceSeer. Different from other face
recognition platforms, FaceSeer contains advanced cognitve
closed-loop feedback to enable the human ability of face recall
into computers. To recall a face is not a simple image processing
problem, it is a complex cognitive task with complex back and
forth oscillation between cognition and visual perception.
FaceSeer organizes the cognitive components of human faces by
using physical linguistics and makes it reliable to match partial
faces against the database. (January 25, 2005, Tucson, Arizona,
USA.)
How it works
FaceSeer uses a comprehensive cognitive model of
human faces to guide its image processing modules. This cognitive
model, which can be treated as a nonlinear functional network, is
built based on physical linguistics. When a face feature presents,
the cognitive model will evoke itself with different constraints
in computational verbs and computational nouns. In turn, the
physical features and measures in the image make the flexible
constraints in the cognitive model become stiff thus finish a
matching process of cognitive features. The measures of the
offsets for each computational verb and noun in the cognitive
model function as the "features" in conventional face recognition
algorithms. A subject face is recognized only when the cognitive
model fits a big enough similarity onto it against some stored
face images.
How to install
Simple, just click the installer icon and you are
ready to go! You can import you face database or if you are
building your face database from scratch, you can collect face
images by using video cameras. FaceSeer also provides function to
organize face databases based on images.
Capacities and Advantages
Can do real-time face recognition with a
640x480 video camera. The face tracking and face locating module
also available with special order.Since the cognitive model of a
face is both global and local, consists both contour of the entire
face and the shapes of eyes, nose, lips and chin, the attention
focus of FaceSeer can freely zoom into the lowest level of lips
and eye lids as well as zoon out to the configuration of entire
face and hair lines. With this cognitive framework, FaceSeer provides a capacity to recognize a person with partial face
images.
Screen shots of the software
Case 1: When the upper half of the face is used to inquiry the
face database. The center panel shows the real image containing
the faces to be recognized. This image can be a frame of a video
camera or from an image file. The region within the white
rectangle is the region used to inquiry the face database. The
right-hand side panel shows a portion of the face database. The
left-hand side panel shows the three best matches from the face
database.
Case 2: When the left half of the face is used to inquiry the face
database. The center panel shows the real image containing the
faces to be recognized. This image can be a frame of a video
camera or from an image file. The region within the white
rectangle is the region used to inquiry the face database. The
right-hand side panel shows a portion of the face database. The
left-hand side panel shows the three best matches from the face
database.
