Gaze Estimation based on Visual Attention Model

Abstract

Gaze information about human is important to judge the attention function in the human. For example, gaze information during car driving and cooking are important. We suggested being able to evaluate attention function from activities of daily living. We acquire gaze information using an eye tracker. The eye tracker is able to estimate high accuracy, however, it is expensive and difficult to operate. In this paper, we discuss a method to estimate gaze without calibration. Specifically, we estimate gaze from visual attention model inhuman. The visual attention model is made from a bottom-up model and a top-down model. However, the brain in human has parts that react to face, body-parts and so. Therefore, we defined a category-specific attention, integrates previous model and category-specific attention, and estimate gaze information. As a result, we present that gaze is similar using the proposed model than the previous model.