Analyzing Evolutionary Art Audience Interaction by Means of a Kinect Based Non-intrusive Method

This paper analyzes the perception by the audience of Evolutionary Works of Art which were produced by means of the unplugged evolutionary algorithm. The long term goal is to study if genetic operations applied by artists in an evolutionary art work are consistently understood by the audience visiting the art exhibit. Yet, we need to involve the audience in the experiment, so that enough data can be retrieved and compared with the way the artists work. Thus, by means of a series of experiments that took place in an art exhibit environment, we analyze audience behavior using two different approaches, the more traditional survey based approach and a new non-intrusive methodology which relies on a depth camera. We show how the latter can largely increase the amount of data collected. Furthermore, it allows to measure both the time spent by a given person in front of an art work, but also to collect additional features about his face, where he is looking and basic gesture recognition. These features can be used to predict their personal attitude and feelings when facing a given art work. Although we describe here preliminary results, they allow us to conclude the pertinence of the approach.