Structural Analysis Of Printed Mathematical Expressions

With the information epoch coming, we need more and more information for our work and life. The input of the information into computer has become a problem and make against the spread of information. The OCR (Optical Characters Recognition) system is a good solution to input quickly and automatically document information into computer and avoid the shortcoming of wasting time and energy by manual. However, there is no commercial OCR system can recognize the content of scientific document including mathematical expressions for its complexity of component and the two-dimension structure, which prevent the scientific document from being digitized. Therefore, automated recognition of mathematical expressions is a challenging pattern recognition problem of great practical importanceIn this paper, we will begin to study structural analysis of mathematical expressions that is a main step in mathematical expressions recognition. Firstly, we divide the symbol into several categories according to its peripheries characters; Secondly, we label the spatial relations between the symbols that are horizontal adjacency on the dominated baseline and the symbols on the nested baselines to build the prime baseline structure tree, at the same time we will partition symbols into units; Next, in order to get the logic relationship among symbols , we begin to do expression syntax analysis that combining LL(1) strategy in context-free grammar with the symbol domain, and do expression semantic analysis that relying on operator domain and advanced processing; Finally, the baseline structure tree is translated into an operator tree according to the precedence of operators. This strategy allows us to exploit the left-to-right reading order of mathematical notations, thereby analyzing layout efficiently without backtracking. In experiments, this method has shown favorable adaptability for the structure of typical mathematical expressions.