Analysing User Visual Implicit Feedback in Enhanced TV Scenarios

In this paper, we report on user behaviors by analyzing visual clues while users are watching various TV broadcast in pilot settings. We detail the first results of the empathic analysis of viewers watching four distinct videos in dedicated recording sessions. Viewers are sitting in front of a TV set in unconstrained position (free postures, free head poses and free body movements) on a chair and recorded by a regular webcam at both low and high resolutions. We have extracted metrics related to: head and global movement, changes in head orientation and facial expressions (happy, angry, surprise). We have conducted preliminary studies about how the extracted metrics can be employed in order to detect the interest, the amusement or the distraction of a viewer.