Towards Culture-Aware Co-Speech Gestures for Social Robots

Embedding social robots with the capability of accompanying their sentences with natural gestures may be the key to increasing their acceptability and their usage in real contexts. However, the definition of natural communicative gestures may not be trivial, since it strictly depends on the culture of the person interacting with the robot. The proposed work investigates the possibility of generating culture-dependent communicative gestures, by proposing an integrated approach based on a custom dataset composed exclusively of persons belonging to the same culture, an adversarial generation module based on speech audio features, a voice conversion module to manage the multi-person dataset, and a 2D-to-3D mapping module for generating three-dimensional gestures. The approach has eventually been implemented and tested with the humanoid robot Pepper. Preliminary results, obtained through a statistical analysis of the evaluations made by human participants identifying themselves as belonging to different cultures, are discussed.