基于机器学习的手写数字识别系统设计与实现  

Design and Implementation of Handwritten Digital Recognition System Based on Machine Learning

作　　者：李怡轩[1] LI Yi-xuan(Weinan Normal University,Weinan 714000)

机构地区：[1]渭南师范学院,渭南714000

出　　处：《微型电脑应用》2018年第8期78-81,共4页Microcomputer Applications

摘　　要：针对传统的手写数字识别准确率低的缺点,将机器学习方法引入手写数字识别。提取数字图像的水平交点、垂直交点和对角交点作为手写数字图像的特征向量,建立手写数字模板矩阵,通过计算待识别图像和模板矩阵的欧式距离和后验概率,从而实现手写数字识别。研究结果表明,机器学习方法手写数字识别的精度可以高达97.63%,为手写数字识别提供新的方法和途径。In view of the low accuracy of traditional handwritten numeral recognition, the machine learning method is intro- duced. The horizontal intersection, vertical intersection and diagonal intersection of digital images are extracted as feature vec- tors of handwritten numeral images, and a handwritten numeral template matrix is established. The handwritten numeral rec- ognition is realized by calculating the continental distance and post-test probability of the images and the template matrix. The results show that the precision of handwritten digital recognition can be as high as 97.63%, which provides a new way for handwritten digital recognition.

关 键 词：机器学习 手写数字 贝叶斯分类器 欧式距离 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]