Why not change the world !


Jixu Chen


Home | Research  | Publications | Demos| Resume| 

               Probabilistic Gaze Estimation Without Active Personal Calibration

(CVPR paper)

Jixu Chen, Qiang Ji
Department of Electrical, Computer and System Engineering
Rensselaer Polytechnic Institute

Existing eye gaze tracking systems typically require an explicit personal calibration process in order to estimate certain person-specific eye parameters. For natural human computer interaction, such a personal calibration is often cumbersome and unnatural.

We propose a new probabilistic eye gaze tracking system without explicit personal calibration:

  • Our approach estimates the probability distributions of the eye parameter and the eye gaze, by combining image saliency with the 3D eye model.
  • By using an incremental learning framework, the subject doesn’t need personal calibration before using the system. His/her eye parameter and gaze estimation can be improved gradually when he/she is naturally viewing a sequence of images on the screen.
  • The proposed system can achieve less than three degrees accuracy for different people without calibration.


Comparison between the traditional gaze estimation and the probabilistic gaze estimation :

 Figure 1. Diagram of traditional gaze estimation

o: optical axis estimated from eye image;

g: gaze point (user's focus) on the screen;

k: person-specific parameter of 3D eye-ball model which determines the relationship between g and o.

In traditional gaze estimation system, the user is asked to look at some specific calibration points on the screen (g*). His/her eye parameter (k*)  is estimated in this calibration procedure. 


 Figure 2. Diagram of probabilistic gaze estimation

p(g|I): probability of the gaze position given the image shown on the screen. It is computed as a saliency map.

p*(k): Probability of the eye parameter estimated from saliency map. It is computed through incremental learning.

p(g|I,o): probability map of gaze given the shown image and the optical axis.


In probabilistic gaze estimation , our system can automatically update the eye parameter probability and estimate the gaze when the user is naturally looking at the screen.  Our approach doesn't need the cumbersome and unnatural personal calibration. (Please check our paper for more details)