Concepture: a regular language based framework for recognizing gestures with varying and repetitive patterns

  • Authors:
  • N. Donmez;K. Singh

  • Affiliations:
  • University of Toronto;University of Toronto

  • Venue:
  • Proceedings of the International Symposium on Sketch-Based Interfaces and Modeling
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

We present Concepture, a framework based on regular language grammars for the authoring and recognition of sketched gestures with infinitely varying and repetitive patterns. Such gestures, while often seen in gesture based applications are currently hard-coded and not customizable. We endorse an example-based workflow, where users author gestures by sketching one or more example instances of the gesture. We de-construct these examples into perceptible stroke segments. Adjacent segment-pairs further capture local spatial relationships between segments and these segment-pairs form the alphabet of a regular language. We then initialize a grammar for our gesture by admitting strings that represent the user provided examples. Grammar refinement is user-friendly, in that we automatically generate new candidate gestures that are visually presented to the user for verification as instances of the gesture. We show Concepture to be effective in efficiently authoring a number of common, yet difficult to recognize gestures, and illustrate it using clip-art and image annotation applications.