Touch Recognition on Complex 3D Printed Surfaces using Filter Response Analysis

Touch sensing on various surfaces has played a prominent role in human-computer interaction in the last decades. However, current technologies are mostly suited for flat or sufficiently smooth surfaces and touch sensing on complex geometries remains a challenging task, especially when the sensing hardware needs to be embedded into the interactive object. In this paper, we introduce a novel sensing approach based on the observation that conductive materials and the user’s hand or finger can be considered a complex filter system with well-conditioned input-output relationships. Different hand postures can be disambiguated by mapping the response of these filters using an intentionally small convolutional neural network. Our experiments show that even straight-forward electrode geometries provided by common 3D printers and filaments can be used to achieve high accuracy, rendering expressive interactions with complex 3D shapes possible while allowing to integrate the touch surface directly into the interactive object. Ultimately, our low-cost and versatile sensing approach enables rich interaction on a variety of objects and surfaces which is demonstrated through a series of illustrative experiments.