青岛大学学报（自然科学版）2024，Vol.37Issue(1)：45-51,7.DOI:10.3969/j.issn.1006-1037.2024.01.08
基于Swin Transformer和CNN的汉字书法教学系统
Chinese Character Calligraphy Teaching System Based on Swin Transformer and CNN
摘要
Abstract
In response to the growing demand for Chinese calligraphy learning,a model combining the Swin Transformer(ST)and Convolutional Neural Network(CNN)was proposed for handwritten Chinese char-acter recognition,subsequently leading to the development of a Chinese character calligraphy teaching sys-tem.The system employed an ST-CNN model for handwriting recognition and classification.The experi-mental results show that the recognition accuracy of the proposed ST-CNN model is around 91.6%,which has a 0.5 percentage points improvement over the traditional ST model.Moreover,the convergence speed of ST-CNN has been improved by about 10 and 30 percentage points respectively compared with traditional CNN and ST models.The developed calligraphy teaching system demonstrates good stability and perform-ance.关键词
深度学习/滑动窗口自注意力模型/卷积神经网络/手写体汉字识别Key words
deep learning/swin transformer model/CNN/handwritten Chinese character recognition分类
信息技术与安全科学引用本文复制引用
林粤伟,张通,宋丹,梁汇鑫,薛克程..基于Swin Transformer和CNN的汉字书法教学系统[J].青岛大学学报（自然科学版）,2024,37(1):45-51,7.基金项目
青岛科技大学公派访学项目资助 （）
2022年青岛科技大学教学改革研究面上项目(批准号:2022MS045)资助 （批准号:2022MS045）
2023年国家级大学生创新创业训练计划项目(批准号:202310426214)资助. （批准号:202310426214）
