On-Line Handwritten flowchart Recognition, Beautification and Editing System

Abstract

In order to segment and recognize on-line handwritten flowchart symbols precisely, we propose a method that segments the graphic symbols based on the loop structure and recognize the segmented symbols by using SVMs. In our experiments, low error rate of 3.37% for symbol segmentation and high recognition rate of 97.6% were obtained. We also propose a beautification and editing method for recognized symbols, and implement them to construct a prototype system. We compare an input time for drawing flowcharts between our system and a traditional application using icon-based interface. As a result, the input time on our system was faster than that on traditional one for flowcharts without texts.