A study of symbol segmentation method for handwritten mathematical formula recognition using mathematical structure information

Symbol segmentation is very important in handwritten mathematical formula recognition, since it is the very first portion of the recognition process. This paper proposes a new symbol segmentation method using mathematical structure information. The base technique of symbol segmentation employed in the existing methods is dynamic programming which optimizes the overall results of individual symbol recognition. The new method we propose here improves symbol recognition performance by using correction values together with evaluation values of symbol recognition. These correction values are calculated from the relations among handwritten stroke positions and mathematical structure. There is no report which takes account of mathematical structure information for symbol segmentation in the handwritten mathematical formula recognition. Our experiments have proven that the recognition rate of symbol segmentation by existing methods is between 90.2% and 93.3%, while our proposed method gives correct recognition rate of 97.1%.