Ecological Gestures for HRI: the GEE Corpus

Ecological Gestures for HRI : the GEE Corpus Maxence Girard-Rivier, Romain Magnani, Véronique Aubergé, Yuko Sasa, Liliya Tsvetanova, Frédéric Aman, Clarisse Bayol. LIG-Lab, University of Grenoble Alps, France E-mail: name.surname@imag.fr Abstract As part of a human-robot interaction project, we are interested by gestural modality as one of many ways to communicate. In order to develop a relevant gesture recognition system associated to a smart home butler robot. Our methodology is based on an IQ game- like Wizard of Oz experiment to collect spontaneous and implicitly produced gestures in an ecological context. During the experiment, the subject has to use non-verbal cues (i.e. gestures) to interact with a robot that is the referee. The subject is unaware that his gestures will be the focus of our study. In the second part of the experiment, we asked the subjects to do the gestures he had produced in the experiment, those are the explicit gestures. The implicit gestures are compared with explicitly produced ones to determine a relevant ontology. This preliminary qualitative analysis will be the base to build a big data corpus in order to optimize acceptance of the gesture dictionary in coherence with the “socio-affective glue” dynamics. Keywords: Human-Robot Interaction, gestures, gesture recognition, “socio-affective glue”, Wizard of Oz experiment 1. Introduction In face-to-face language interactions, the facial expressions, the body gestures and proxemics have been explored but we usually notice them in a wider range of body language studies (Scheflen, 1972; Gallagher, 2005). Gestures are usually considered as a paralinguistic feature but in sign languages, they carry all the complexity of language and interactions (Kendon, 1994). In Human-Robot Interaction (HRI), the human gestuality has been studied mainly as complementary information to speech. This work is a part of the Interabot (Investissements d’Avenir) Project, held with the robotics Awabot Company developing the Emox robot and the LIRIS Laboratory where an efficient DNN gestures recognition system is developed. One main goal of this presented study is to demonstrate that human gestures produced naturally without focus on the gestural commands are very different from gestures they would have proposed if they would have been explicitely asked to produce gestures for HRI (as for example in “focus groups” as proposed in ergonomy methods). Thus, the scenario implying the human in a quite ecological situation, gives only the gestural modality to communicate with a robot. If the basic gestures of this corpus of Gestural Emoz Expressions (GEE) are possible to be learned by automatic gestures recognition systems, that could be a way to introduce HRI without creating and teaching artifactual gestures to the human user. The corpus collected for this study is set in an ecological micro- world, as we need to observe gestures spontaneously produced by humans in a real HRI situation (Guillaume & al, 2015). As a consequence, to train an HRI system to natural gestures should not impose a gesture language to the human. This could directly optimize the acceptance of technology, and to ensure its durability. These gestures, when validated, can be …

Content truncated to 500 words