An Improved Two-Dimensional Gaze Mapping Model

This paper presents an improved two-dimensional gaze mapping model based on active infrared light source. The concerned parameters of gaze were detected under the status of static head and the interpupillary distance was compensated. Then a model was built to map the gaze parameters to the gaze point using nonlinear polynomial. At last the technique of support vector regression (SVR) was used to compensate the deviation of sight line caused by different head position, so that non-linear mapping function can be extended to any head position. The results of the experiments verify that the improved model is valid and feasible.