基于卷积循环神经网络的手写汉字文本识别
DOI:
作者:
作者单位:

河北工程大学机械与装备工程学院

作者简介:

通讯作者:

中图分类号:

TP391.1

基金项目:

河北省教育厅科学研究项目(CXY2024046)


Handwritten Chinese Character Text Recognition Based on Convolutional Recurrent Neural Network
Author:
Affiliation:

Institute of Machinery and Equipment Engineering

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    为了解决卷积循环神经网络(CRNN)手写汉字文本识别网络模型的训练参数大、文本识别率低等问题,提出了一种基于注意力双向长短期记忆网络(AT-BLSTM)和知识蒸馏(KD)技术的手写汉字识别方法。通过对AT-BLSTM网络的输入向量特征赋予不同的权重,使模型训练数据集更加高效、准确;通过KD技术将一个高性能的大模型获取的知识传输到一个小模型中,在确保模型准确性的同时,减少训练参数和内存占比,得到一个性能更优的轻量级训练模型。该方法通过多组实验对比,汉字识别准确率提高了6.7%,训练参数减少15.94M。该网络模型识别准确率达到了97.9%,汉字识别效果更好。

    Abstract:

    In order to solve the problems of large training parameters and low text recognition rate of convolutional recurrent neural network (CRNN) handwritten Chinese character recognition network model, a novel method for handwritten Chinese character recognition based on attention bi-directional long short-term memory network(AT-BLSTM) and knowledge distillation (KD) technology is proposed. By assigning different weights to the input vector features of AT-BLSTM network, the model training data set is more efficient and accurate. Through KD technology, the knowledge acquired from a large high-performance model is transferred to a small model, which ensures the accuracy of the model, reduces the training parameters and internal storage ratio, and obtains a lightweight training model with better performance. Through the comparison of multiple groups of experiments, the accuracy of Chinese character recognition is increased by 6.7%, and the training parameters are reduced by 15.94M. The recognition accuracy of this network model reaches 97.9%, and the recognition effect of Chinese characters is better.

    参考文献
    相似文献
    引证文献
引用本文

胡瑞朋,何春燕. 基于卷积循环神经网络的手写汉字文本识别[J]. 科学技术与工程, , ():

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2024-04-24
  • 最后修改日期:2024-05-21
  • 录用日期:2024-05-22
  • 在线发布日期:
  • 出版日期:
×
亟待确认版面费归属稿件,敬请作者关注