基于卷积神经网络的手写汉字识别方法  

Recognition Method of Handwritten Chinese Characters Based on Convolutional Neural Network

作　　者：肖婷婷 XIAO Tingting

机构地区：[1]华东理工大学,上海201424

出　　处：《科技创新与应用》2021年第25期114-115,118,共3页Technology Innovation and Application

摘　　要：手写汉字识别在智能化处理手写文档时至关重要,高准确率的预测结果对后续处理文本信息有着重要作用。为解决传统汉字识别特征提取繁琐等问题,使用深度学习中的卷积神经网络,利用Keras搭建网络离线识别手写汉字图片。在HWDB1.1的数据集上,构建6层小网络,预测图片中的汉字。分析实验结果得知,深度学习可以有效提取图片特征,tanh,sigmoid,relu三个激活函数中relu效果最好。Handwritten Chinese character recognition is very important in the intelligent processing of handwritten documents,and the high accuracy prediction results play an important role in the subsequent processing of text information.In order to solve the problem of tedious feature extraction in traditional Chinese character recognition,convolutional neural network(CNN)in deep learning is used,and Keras is used to build a network for off-line recognition of handwritten Chinese characters.On the data set of HWDB1.1,a 6-layer small network is constructed to predict the Chinese characters in the image.Analysis of the experimental results show that deep learning can effectively extract image features,and relu is the best among the three activation functions of tanh,sigmoid,relu.

关 键 词：HCCR 卷积神经网络 HWDB数据集 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]