Research About Designing Of A Typeset Mathematical Expression Understanding System

Mathematical expressions constitute an essential part in most scientific and engineering documents. With rapid development of modern internet technology and computer technology, computer have penetrated in all kinds of society life domains, and Human being have step into information era. There is a key vehicle of disseminating and exchanging information via the Internet. It is important to the researching of us transcribing the scientific and engineering documents into electronic form. When we transcribe existing knowledge in the form of paper documents into corresponding electronic form, a lot of information that we have possessed can be processed by today's digital computer and transmitted through the Internet. The recognition of mathematical expressions is becoming a key in transcribing documents in scientific and engineering disciplines into electronic form. In this paper, we discuss recognition of mathematical expressions in the scientific and engineering documents. First, reviewing the history of the recognition of mathematical expressions and analyzing the properties of mathematical expressions. Then we can divide the mathematics-recognition problem into three processes: page segmentation and symbol segmentation , symbol recognition, structural analysis. We detected the mathematical expressions printed in separate lines by calculating the y coordinate of the bottom-most row of symbol and it's standard deviation. For detecting embedded mathematical expressions, we can check each line to find one or more of the mathematical symbols we define. The segmentation of symbol can be done by recursive horizontal and vertical projection profiles cutting. We employ a flood fill algorithm deal with the sub-expression which cannot be deal with by the projection profile cut. The Support Vector Machines (SVM) is a important application of the statistical learning theory. We apply it to the symbol recognition and it express a good performance. In the processes of structural analysis, we use tree transformation to analysis the structure of mathematical expressions. We introduce the conception of Baseline Structure Tree (BST), the internal tree nodes are operators and leaf nodes are operands. The way of tree transformations can express the structure of mathematical expressions in a convenient and compact form. In this paper, we give a survey of recognition of mathematical expressions, and reviewing a lot of way in the three processes of recognition. We give the programs and results of recognition of mathematical expressions in print documents. Finally, we discuss the difficulty about recognition of mathematical expressions and it's trend in the future.