手写字符的快速识别模型  

A fast recognition model for handwritten characters

作　　者：刘淑明 巩荣芬[1] 储茂祥[1] 杨永辉[1] LIU Shuming;GONG Rongfen;CHU Maoxiang;YANG Yonghui(School of Electronic and Information Engineering,University of Science and Technology Liaoning,Anshan 114051,China)

机构地区：[1]辽宁科技大学电子与信息工程学院,辽宁鞍山114501

出　　处：《辽宁科技大学学报》2022年第6期438-442,共5页Journal of University of Science and Technology Liaoning

基　　金：国家自然科学基金(71771112);辽宁省自然科学基金(2022-MS-353);辽宁省教育厅基本科研面上项目(LJKMZ20220640)。

摘　　要：现有字符识别模型识别手写字符速度慢效率低。本文提出一种手写字符的快速识别模型,采用双通路卷积特征提取方式,一路分支采用常规卷积神经网络,另一路分支采用空洞卷积神经网络。不同分支的卷积神经网络能够更好地获取不同感受野的手写字符特征。提出一种图像处理方式,能够更好提取随机分布的手写字符。实验结果表明,该模型不仅拥有高识别率,同时在速度上拥有绝对优势。The existing character recognition models are slow and inefficient in recognizing handwritten characters. This paper proposes a fast recognition model for handwritten characters. The binary branch convolution feature extraction method is adopted. One branch uses conventional convolution neural network,and the other branch uses dilated convolution neural network. The convolutional neural networks of the two branches can better obtain the handwritten character features in different receptive fields. An image processing method is proposed to better extract randomly distributed handwritten characters. The experimental results show that the model has not only high recognition rate,but also absolute advantages in speed.

关 键 词：手写字符识别 深度学习 卷积神经网络 空洞卷积神经网络 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]