BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Eye Tracking with Consumer Hardware - Dan Witzner Hansen\, IT Univ
 ersity of Copenhagen
DTSTART:20060111T150000Z
DTEND:20060111T160000Z
UID:TALK4650@talks.cam.ac.uk
CONTACT:Phil Cowans
DESCRIPTION:Present commercial gaze trackers (i.e. from Tobii and LC\nTech
 nology) are easy to use\, robust and sufficiently\naccurate for many scree
 n-based applications but their costs\nexceed the budget of most people. Lo
 w cost eye tracking has\nreceived an increased attention due to the rapid\
 ndevelopments in tracking hardware (video boards\, digital\ncamera and CPU
 s). Eye tracking based on consumer hardware\nis subject to several unknown
  factors as various system\nparameters (i.e. camera parameters and geometr
 y) are\nunknown. Robust statistical principles to accommodate\nuncertainti
 es in image data are therefore needed. I will\ndiscuss the components (det
 ection\, tracking and gaze\nestimation) used in a low-cost eye tracker. I 
 will in\nparticular describe our contour-based iris tracker. The\ncontour 
 model is based on the statistics of natural images.\nIt turns out that thr
 ough fairly simple modeling that\nexplicit feature detection can be avoide
 d and thus\nthresholds become needless. Based on the data from the eye\ntr
 acker I will then discuss current gaze estimation methods\nand compare the
 m with gaze estimation methods using\nGaussian Processes.
LOCATION:Ryle Seminar Room\, Cavendish Laboratory
END:VEVENT
END:VCALENDAR