Accurate Annotation of Gaze Point for Wearable Eye Tracking Devices

Chao Liang,Jing Chen,Zhongmin Cai

DOI: https://doi.org/10.1109/cac53003.2021.9728046

2021-01-01

Abstract:As the cost of eye tracking devices decreases, wearable eye tracking devices are becoming more and more popular in head-mounted devices such as smart glasses and augmented reality helmets. Eye tracking enabled devices are believed to facilitate a better understanding of human behaviors and intentions, and will make the interaction between human and computing systems more natural and friendly. A challenging problem for wearable eye tracking devices is the accurate annotation of the gaze point in a real-world scenario, that is, to identity the real object where each gaze point is on and the accurate position of the gaze point relative to this object. We propose a framework for gaze point annotation in real-world scenarios and design a system to implement this framework by establishing an accurate mapping between the gaze point and its coordinates relative to a computer screen. Experimental results show our method is robust in real-world scenarios and can reach an error rate less than 1%, This demonstrates the effectiveness of the proposed framework.