A Symbolic Approach for Task-related Gaze Classification

Abstract

This paper introduces an innovative symbolic framework for task-related gaze classification. The proposed method encodes multivariate eye-tracking feature signals into symbolic string sequences, facilitating the computation of an interpretable and robust distance metric between recordings. Applied to the ETRA 2019 dataset, the symbolic representation enables accurate clustering of visual tasks without dependence on black-box models. Experimental results demonstrate that our approach achieves performance comparable with state-of-the-art deep learning methods.