Recognition of Multi-touch Drawn Sketches

Michael Schmidt, Gerhard Weber

Abstract: We present concepts and possible realizations for the classification of multi-touch drawn sketches. A gesture classifier is modified and integrated into a sketching tool. The applied routines are highly scalable and provide the possibilities of domain independent sketching. Classification rates are feasible without exploiting the full potential of the scheme. We demonstrate that the classifier is capable of identifying common basic primitives and gestures as well as complex drawings. Users define sketches per templates in their individual style and link them to constructed primitives. A pilot evaluation is conducted and results regarding sketching techniques of users and classification rates are discussed.