基于SVM的印刷体数学公式识别的研究
Research on Recognition of Printed Mathematical Formula Based on SVM

传统的数学公式识别，通常建立在OCR技术进行图片文字识别的基础上，对目标公式进行符号切割，通过构建数学符号数据库，然后两两比较相似度，然后返回最大相似度的符号名称，作为识别结果。该方法，对数学符号数据库要求极高，鉴于实际情况，公式存在字号大小、粗细体、正斜体、各种字体等差异，导致该方法识别效果不佳。本文基于印刷体数学公式特点，重新构建字符标准库，并结合机器学习思想，应用SVM算法进行公式识别，并进一步提取字符特征，提升公式识别精度，实验结果显示，识别结果良好。
Traditional mathematical formula recognition, usually based on OCR technology for image and text recognition, cuts the symbol of the target formula, builds the mathematical symbol database, com-pares the similarity, and then returns the symbol name of the maximum similarity as the recogni-tion result. In view of the actual situation, there are some differences in the formula, such as font size, thickness, italics, various fonts and so on. Based on the characteristics of printed mathematical formulas, this paper reconstructs the character standard library, and combines with the machine learning idea, uses SVM algorithm to recognize formulas, and further extracts the character features, improves the accuracy of formula recognition. The experimental results show that the recognition results are good.