Gaze detection by estimating the depths and 3D motion of facial features in monocular images

摘要

This paper describes a new method for detect- ing the gazeposition of a user on a monitor from monocular images. In order todetect the gaze position, we extract facial features(both eyes,nostrils and lip corners)automatically in 2D camera images andestimate the 3D depth information and the initial 3D positions ofthose features by recursive estimation al- gorithm in startingimages. Then, when a monitor, the moved 3D positions of thosefeatures can be estimated from 3D motion estimation by ExtendedKalman Filter(EKF)and affine trans- form.