基于深度学习和表意文字描述序列的多种类汉字识别方法
By using deep learning and ideographic character description sequences, a large number of data samples were generated and a recognition network was trained, which solved the problem of recognizing rare characters and characters in the official script. This enabled accurate recognition of rare characters and differentiation of characters in the official script, thus improving the accuracy of ancient character recognition.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- HUAZHONG UNIV OF SCI & TECH
- Filing Date
- 2023-10-07
- Publication Date
- 2026-07-17
AI Technical Summary
Existing character recognition methods struggle to accurately identify rare characters and distinguish between characters with established script, especially in the study of ancient characters, where current technology cannot effectively address the problem of recognizing and differentiating between rare characters and characters with established script.
This study employs a deep learning-based approach using ideographic character description sequences. By generating a large number of data samples, including images of non-existent Chinese characters, the recognition network is trained. The model is then optimized using residual networks and cross-entropy loss functions, thereby improving the accuracy of recognizing rare characters and enhancing the ability to distinguish between characters from the official script.
It significantly improves the variety and accuracy of Chinese character recognition, especially the recognition rate of rare characters, and can accurately distinguish clerical script characters, thus improving the accuracy of ancient character recognition.
Smart Images
Figure CN117333883B_ABST