Effects of tracker location on the accuracy and precision of the Gazepoint GP3 HD for spectacle wearers.

Abstract

Remote view eye-tracking systems are prone to errors when used on spectacle wearers due to reflections from the lenses and frame that result in inaccurate tracking. Traditionally, these trackers are situated below a computer monitor and the viewer's eye moments are recorded while they view the screen. Reflections may be influenced by the pantoscopic tilt of the spectacles, whereby the tilt angle causes incident light to be reflected to the camera. To overcome this problem, we propose mounting the tracker above the monitor to avoid these reflections and test the accuracy and precision of subjects with single vision spectacles, multifocals, and no correction, using both mounting positions. Experimental results showed that this alternate position had overall worse accuracy (4.06° ± 0.13) and precision (0.67° ± 0.05) compared to the standard configuration (2.15° ± 0.06 vs. 0.50° ± 0.03), with more invalid readings (5.91 vs. 19.19%) for single vision lens wearers. Multifocals performed better for the top-mounting position for the top portion of the monitor, suggesting higher-order aberrations from the bottom portion of the lens negatively impact data quality. Higher pantoscopic tilt angles displayed an improved accuracy for this alternate position (r(9) = - 0.69, p = 0.02), with superior accuracy for tilt angles greater than 14° compared to the standard configuration. This study quantifies the impact of spectacle wear on eye-tracking performance and suggests other alternate mounting positions may be viable in certain situations.