Segmentation And Recognition System Based On Neural Network Mathematical Formula Symbol

The neural network has a history of over fifty years now. Generations of researchers have been making great efforts to build up its theoretical foundation and to apply it in many areas such as signal processing, machine vision, pattern recognition, expert system, industry control and weather forecast. In recent years, the pattern recognition based on neural networks has become a new active field. The study of the neural networks-based pattern recognition system is very important, not only to the development of neural networks theories, but also to the application of the pattern recognition techniques.The multi-layer feed-forward back-propagation neural network has found a widely expanding range of applications. It has been used extensively on OCRThe SOFM proposed by Kohonen is also used in the area of pattern recognition for its strong organizing ability on topology and its robustness.The work we have done is mainly focused on the following two problems.First, the generalized inverse matrix has many important applications such as neural network computing. But at present there is no clear and easy-to-understand treatment on it in our country. In paper[3],the generalized inverse matrix is defined by abstract algebra method, which is difficult to understand and to use. In Chapter 2 of this paper, we try to give a comprehensive treatment of the definition and properties of generalized inverse matrix by use of some simple and easy-to-understand tools such as matrices and inner products. A discussion on its application to neural networks is also provided.Secondly, as mentioned above, the neural networks is has been used extensively on OCR. But the problem is that many researchers work under the hypothesis that the characters were segmented perfectly. What's more, few papers specially discuss the recognition of symbols in mathematical expressions based on the neural networks. Automatic recognition of mathematical expressions is one of the key vehicles in the drive towards transcribing documents in scientific and engineering literatures into electronic form.In this paper, the segmentation and the recognition of mathematical expressions are considered, based on the BP neural networks, SOFM network and moment technique. It consists of four major stages: pretreatment, symbol segmentation, symbol recognition and structural analysis.In Chapter 3, the segmentation of the mathematical expressions is proceeded after the pretreatment. We use three segmentation methods according to various features of the symbols.This paper is focused on the special symbols set------mathematicalexpression which has 134 patterns. So a multi-stage neural networks (MSNN) is adopted. In order to account for the recognition of mathematical symbols with translation, rotation and scaling in a certain range, we preprocess by moment technique a great number of examples to choose a final set of representative training set. And because the recognition is linked with segmentation, the system proves to be more practical.Due to the limitation of the examples selected and the algorithms used, the model we proposed still needs further improvement for practical application. But the results have shown that the neural networks can work successfully on the recognition of mathematical symbols.