This paper reports a novel stereo-vision-method (binocular system-geometrical mapped (BS-GM)) to estimate the depth coordinates of the eye gaze point in a controlled 3D space of vision. The method outcomes were compared in both 2D and 3D visual targets with both mono- and stereovision algorithms in order to estimate accuracy of results. More specifically, we compared BS-GM with a monocular method and with two stereo-vision methodologies which were different in order to the mapping functions. All of the methods were implemented in the same head mounted eye tracking system able to acquire both eyes. In 2D visual space (i.e. plane of vision) we compared BS-GM with a monocular method, a binocular system-linear mapped (BS-LM) and a binocular system-quadratic mapped (BS-QM). In the 3D space estimation all of the binocular systems were compared each other. Thirteen enrolled subjects observed 31 targets of known coordinates in a controlled environment. Results achieved on 2D comparison showed no statistical significant difference among the four methods, while the comparison on 3D space of vision showed that BS-GM method achieved a significant better accuracy than BS-LM and BS-QM method. Specifically, BS-GM showed and average percentage error obit of 3.47%.
On the tridimensional estimation of the gaze point by a stereoscopic wearable eye tracker
LANATA', ANTONIO;GRECO, ALBERTO;VALENZA, GAETANO;SCILINGO, ENZO PASQUALE
2015-01-01
Abstract
This paper reports a novel stereo-vision-method (binocular system-geometrical mapped (BS-GM)) to estimate the depth coordinates of the eye gaze point in a controlled 3D space of vision. The method outcomes were compared in both 2D and 3D visual targets with both mono- and stereovision algorithms in order to estimate accuracy of results. More specifically, we compared BS-GM with a monocular method and with two stereo-vision methodologies which were different in order to the mapping functions. All of the methods were implemented in the same head mounted eye tracking system able to acquire both eyes. In 2D visual space (i.e. plane of vision) we compared BS-GM with a monocular method, a binocular system-linear mapped (BS-LM) and a binocular system-quadratic mapped (BS-QM). In the 3D space estimation all of the binocular systems were compared each other. Thirteen enrolled subjects observed 31 targets of known coordinates in a controlled environment. Results achieved on 2D comparison showed no statistical significant difference among the four methods, while the comparison on 3D space of vision showed that BS-GM method achieved a significant better accuracy than BS-LM and BS-QM method. Specifically, BS-GM showed and average percentage error obit of 3.47%.I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
