Eye Gaze Calculation Based on Nonlinear Polynomial and Generalized Regression Neural Network

In this paper, we present a method of calculating the direction of the line of sight, which is based on nonlinear polynomial and generalized regression neural network, using an active infrared light source system. First of all, we get a model to map the gaze parameter to the gaze point under the circumstances of a static head with nonlinear polynomial and Generalized regression neural network are then used to compensate for different head position, so that nonlinear mapping function is extended to any head position, and then we can accurately estimate the gaze under head movement. The experiment and application in an interactive graphical interface system have verified the validity of this method.