Recognizing Visual Attention Allocation Patterns Using Gaze Behaviors

Abstract

Human Attention Allocation Strategy (HAAS) is related closely to operating performance when he/she is interacting a machine through a human-machine interface. Gaze behaviors, which is acquisited by eye tracking technology, can be used to observe attention allocation. But the performance-sensitive attention allocation strategy is still hard to measure using gaze cue. In this paper, we attempt to understand visual attention allocation behavior and reveal the relationship between attention allocation strategy and interactive performance in a quantitative manner. By using a novel Multiple-Level Clustering approach, we give some results on probabilistic analysis about interactive performance of HAAS patterns in a simulation platform of thermal-hydraulic process plant. It can be observed that these patterns are sensitive to interactive performance. We conclude that our Multiple-Level Clustering approach can extract efficiently human attention allocation patterns and evaluate interactive performance using gaze movements.