The Attentive Machine: Be Different!

We will demonstrate an intelligent Machine which is capable to choose within a small group of people (typically 3 people) the one it will interact with. Depending on people behavior, this person may change. The participants can thus compete to be chosen by the Machine. We use the Kinect sensor to capture both classical 2D video and depth map of the participants. Video-projection and audio feedback are provided to the participants. 1 Social Feature Extraction The main feature which is extracted is the personal space of the participants. Social studies [1] showed that humans have different “ego-spaces”: the public space (from around 3.5 meters in green on Figure 1, left image), the social space where interaction is possible (from around 1.2 meters in blue on Figure 1, left image), the personal space for close interaction (from around 0.45 meters in yellow on Figure 1, left image) and the intimate space (in red on Figure 1, left image). Those measures vary of course depending on cultural and personal contexts. This space is extracted in 3D [2] by using OpenGL and a Microsoft Kinect sensor (Figure 1, right image). Fig. 1. Right: ego-spaces, Left: 3D extraction of intimate space (red cylinder) and personal space (blue sphere)