Synthetic Word Gesture Generation for Stroke-Based Virtual Keyboards

Gesture typing is a stroke-based text input method for mobile devices. Instead of tapping, the user enters a word by gesturing through all its letters on a virtual keyboard with a single continuous stroke. Even though gesture typing is a widely available text entry method for mobile devices, very little is known about the underlying movement dynamics. The present work fills this gap in the literature by establishing a physiological movement model for gesture typing. We show that many features of word gestures can be described by the mathematical framework of a handwriting model. We extend the model to respect the particularities of gesture typing and show that the exact trajectories as well as the velocity profiles of most word gestures can be represented with a high accuracy. We present an algorithm to extract the model parameters from real gestures. Finally, we introduce a framework capable of synthesizing the trajectories of word gestures that share many features with human-generated gestures. We use these synthetic gestures to automatically evaluate gesture recognition algorithms with thousands of gestures.