Appearance-based Eye Gaze Estimation

  • Authors:
  • Kar-Han Tan;David J. Kriegman;Narendra Ahuja

  • Affiliations:
  • -;-;-

  • Venue:
  • WACV '02 Proceedings of the Sixth IEEE Workshop on Applications of Computer Vision
  • Year:
  • 2002

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present a method for estimating eye gaze direction,which represents a departure from conventional eye gazeestimation methods, the majority of which are based ontracking specific optical phenomena like corneal reflectionand the Purkinje images. We employ an appearance manifoldmodel, but instead of using a densely sampled splineto perform the nearest manifold point query, we retain theoriginal set of sparse appearance samples and use linearinterpolation among a small subset of samples to approximatethe nearest manifold point. The advantage of this approachis that since we are only storing a sparse set of samples,each sample can be a high dimensional vector thatretains more representational accuracy than short vectorsproduced with dimensionality reduction methods. The algorithmwas tested with a set of eye images labelled withground truth point-of-regard coordinates. We have foundthat the algorithm is capable of estimating eye gaze with amean angular error of 0.38 degrees, which is comparableto that obtained by commercially available eye trackers.