Towards Leaning Aware Interaction with Multi-touch Tabletops

Paper in proceeding, 2016

Interactive tabletops allow direct touch manipulation and recognizing simultaneous touch events. Users sometimes lean on the touch surface creating unintended touch input. Our work demonstrates how this unintended input can be employed to enhance interaction. In a study we develop a posture set organized into four classes. We present a vision-based machine-learning algorithm using an active shape model to recognize the classes. The algorithm categorizes lean gestures into one of the classes for interaction purposes. In a second study, we evaluate the model and propose interaction scenarios that use lean detection.