Feature based Methods for Eye Gaze Tracking

Eye gaze has been depicted as one of the means for interaction with computer. The eye tracking systems perform feature extraction and determination on gaze direction. The feature based methods will estimate gaze direction based on the extracted features. The parameters of pupil, iris and eye corners will determine gaze direction. In this paper, two feature based gaze tracking methods are proposed. The first method is based on Fast Fourier Transform (FFT) and the second method is based on number of zeros in Z-domain. The search for template image in successive video frames tracks the pupil. The optical flow extracts the change in position. The velocity component determines displacement of template region, which helps in gaze direction estimation. The Fourier transform is computed by using the magnitude of velocity components, gives the frequency representation for each gaze direction. The number of zeros inside a specific range of circle in Z-domain aids in determining the gaze direction. The recognition rate of 98.5% and 99% are obtained for the methods based on FFT and zeros, respectively.