具有连续反馈的在线草图符号识别算法

A recognition algorithm for online sketched symbols with continuous feedback

当前多数在线草图符号识别方法是在用户绘制完图形后,通过点击按钮或等待一定时间来反馈识别结果。为提供连续反馈,提出了一种基于图元的部分草图识别算法:对输入笔画进行笔画分割得到图元表示,提取表示两个图元空间关系的新特征,依据指派问题模型,通过该特征的相似度识别符号。算法简单,无须训练,而且与笔顺、笔画数和图形大小无关。实验证明了方法的有效性。

At present,clicking on the button and waiting for a period of time after completing the sketch fully is the most common recognition method for online sketched symbols. To achieve the continuous feedback,a primitive-based partial sketches recognition algorithm was proposed. In the stage of feature representation,the input strokes were segmented as primitives and then a new feature which presents the spatial relationship between a pair of primitives was come up with. Based on the model of assignment problem,symbols were recognized through the similarity of this feature finally. The proposed method is so easy that one can master it without training and it has nothing with the order of stroke,the number of stroke and the size of graph. Its validity was testified by experiment.