Recognition efficiency issues for freehand sketches

Sketch understanding has received attention as an enabling technology for natural human-computer interaction (Thomas Stahovich & Randall Davis, 2002). With the widespread availability of pen based PDAs, and more recently with the emergence of Tablet PCs, there is an increasing interest in sketch recognition. Current approaches to sketch recognition treat sketches as static images and apply structural or syntactic recognition techniques commonly used in computer vision. In this paper, we characterize sketching as an interactive, incremental process, and argue that sketch recognition algorithms should be tailored to take advantage of these properties of sketches that separate them from images. We report experimental results showing how the order in which strokes are drawn affects the recognition speed and propose possible approaches for achieving algorithms with better memory and speed requirements.