A Study towards Improving Eye Tracking Calibration Technique Using Support Vector Regression

Visual attention and movement have evolved by many ways in eye tracking system. With the mechanism and functionality of eye movement, tracking human eye with high accuracy is possible. In eye tracking system, calibration technique can be considered as the first crucial step to be taken before eye can be tracked. The technique highly depends on a number of calibration points used and the selection of these points influences the overall performance of eye tracking system. Knowing the possibility, there are various methods of eye tracking with a good calibration process. However, determining the optimum number of calibration points is a huge challenge when developing eye tracking system. The research focus is then oriented towards determining an optimum number of calibration points by integrating the method of Support Vector Regression (SVR) with the conventional calibration technique. The proposed method should be able to reduce processing time while having high eye tracking accuracy. This paper describes the calibration process and analyses the possible optimum number of calibration points which will be able to reduce processing time and improve accuracy of eye tracking system.