Estimation of User's Attention and Awareness in Occlusion-Rich Environments Using RGB-D Cameras

摘要

Objective recognition by systems often does not agree with subjective recognition by users. Therefore, it is an important to estimate users' subjective states appropriately. Especially, in occlusion-rich environments, information on what a user can/cannot see, what he/she pays attention to, and what he/she is aware of or not in the environments is one of important clues to estimate his/her subjective states and predict next actions. In this paper, we propose a system for estimating maps of a user's attention and awareness in such environments based on the view estimation system using RGB-D cameras. The proposed system can estimate what the user sees, what he/she pays attention to, and what he/she is aware of in environments in pixels of captured images. Experimental results in a real environments show effectiveness of the proposed system. Furthermore, we discuss an extension of the proposed system to estimation for multiple users.