Modeling Video Viewing Behaviors for Viewer State Estimation (Authors Version)
Ryo Yonetani · 2012
ACM, (2012). This is the authors version of the work. It is posted here by permission of ACM for your personal use. Not for redistribution. The definitive version was published in the proceeding of ACM Multimedia 2012 Doctoral Sym-posium (ACMMM 2012 DS). Human gaze behaviors when watching videos reflect their cognitive states as well as characteristics of the video scenes being watched. Our goal is to establish a method to esti-mate the viewer states from his/her eye movements toward general videos, such as TV news and commercials. The pro-posed method is based on a novel model of video viewing behaviors, which takes into account structural and statis-tical relationships between video dynamics, gaze dynamics and viewer states. This model realizes statistical learning of gaze information while considering dynamic characteristics of video scenes to achieve viewer-state estimation. In this paper, we present an overview of the viewer-state estima-tion method based on the model of video-viewing behaviors, including several past work done by the author’s team.