A new technique for improving pupil detection algorithm

In this paper are presented several techniques used for improving the pupil detection algorithm based on a head mounted eye tracking system employing an infrared video camera. The coordinates of the eye pupil center obtained by any eye tracker device are affected by many noise sources. In order to improve the pupil detection position, several different techniques of real time filtering and high frequency spikes canceling from the signal provided by the pupil detection algorithm are presented. The optimal parameters of the real time filter, depending on the application sampling frequency, have been determined.