Detecting preferences based on eye movement using combinatorial fusion

Abstract

When tasked with comparing two images on a screen, a subject's eye movement can be captured and analyzed in order to understand the process of preference formation. The process of comparing two images and developing a preference is analyzed based on a sample dataset. Although it is known in general that our preferences are shaped by our past experiences, a systemic understanding of the factors which lead to preference decision making remains a challenging problem. In this paper, we propose a set of five attributes which are extracted from the temporal eye movement sequence: last duration, total duration, gaze count, interest sustainability, and region change. Each of these five attributes is a scoring system (ranking system). We then use the combinatorial fusion algorithm (CFA) framework to combine pairs of attributes using the rank-score characteristic (RSC) function and cognitive diversity (CD). Our results demonstrate that combination of two attributes can improve individual attributes if the attribute pair has a higher cognitive diversity. Our work represents a new paradigm to use combinatorial fusion for preference detection based on eye movement.

Additional details