A Structural Method for Online Sketched Symbol Recognition

The increasing availability of pen-based hardware has resulted in a parallel growth in research in sketch recognition. However, many challenges remain in terms of recognition accuracy, robustness to different drawing styles, rotational invariance, and the number training samples. To address these challenges, a new structural approach to online sketched symbol recognition was proposed, which focuses on the primitive correspondence between a input symbol and the reference one. This method is independent of stroke-order,-number, as well as invariant to scaling and rotation. Experiments on two datasets show the effectiveness of the proposed approach.