Towards a Framework for Evaluating Synthetic Surface Gestures

Synthetic surface gestures are required for efficient test-automation of gesture-based applications, but must resemble human gestures as closely as possible. We thus require options to assess the human-like quality of synthetic surface gestures. In this paper, we discuss several approaches for evaluating synthetic data from other fields and map them to synthetic surface gestures. We also propose a framework to assess a gesture generator's performance and evaluate its feasibility in a preliminary evaluation with three different gestures. Our results indicate that the proposed framework is able to rate different datasets according to expectations of similarity to a baseline dataset.