MatchUp & Conquer

Yosra Rekik, Inria Lille & LIFL, University of Lille1, France

Radu-Daniel Vatavu, University Stefan cel Mare Of Suceava, Romania

Laurent Grisoni, Inria Lille & LIFL, University of Lille1, France

• About

Match-Up & Conquer is a simple, two-step technique for recognizing multi-touch gesture input that is invariant to how users articulate gestures, i.e., by using one or two hands, one or multiple fingers, one or multiple strokes, synchronous or asynchronous stroke input. We introduce, for the first time in the gesture literature, a preprocessing step that is specific to multi-touch gestures (Match-Up) that clusters together similar strokes produced by different fingers, before running a gesture recognizer (Conquer). We report gains in recognition accuracy up to 10% leveraged by our new preprocessing step, which manages to construct a more adequate representation of multi-touch gestures in terms of key strokes. It is our hope that the Match-Up technique will add to the practitioners’ toolkit of gesture preprocessing techniques, as a first step toward filling today’s lack of algorithmic knowledge to process multi-touch input and leading toward the design of more efficient and accurate recognizers for touch surfaces.

• Match-Up & Conquer two step technique is distributed under the LGPL version2 license agreement.

• If you find the code or dataset useful for your work, please let me know. If you use the code or datasets to report results for publications, please reference the work below.

• Multi-touch gestures dataset

The multi-touch gesture dataset contains 22 different gesture types: letters, geometric shapes (triangle, square, horizontal line, circle), symbols (five-point star, spiral, heart, zig-zag), and algebra (step- down, asterisk, null, infinite):