eMir: Digital signs that react to audience emotion

Abstract

In this paper we present eMir, digital signage (public electronic displays) that show human faces which react to audience emotion. Using a camera installed at the sign, the system observes the audience and detects whether someone watches the display via face detection software. The face detection is able to classify facial expressions and determine gender. This information is used to let a human character on the screen react accordingly and encourage interaction with the face/sign. The system has been deployed for one month on a digital sign in a university building. We present experiences with the system, our findings from the collected interaction data and results from interviews with eight users.