用于扫读笔的离线识别方法及系统

By constructing a scanning pen recognition model using deep learning and neural network technologies, the problem of insufficient accuracy of existing scanning pens in recognizing complex paper was solved. It also enabled personalized customization and contextual understanding, expanded the applicability of languages ​​and character sets, and provided more reliable recognition results.

CN117152768BActive Publication Date: 2026-07-17GUANGZHOU SIMWARE TELECOM CO LTD

Patent Information

Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
GUANGZHOU SIMWARE TELECOM CO LTD
Filing Date
2023-08-10
Publication Date
2026-07-17

AI Technical Summary

Technical Problem

Existing scanning pens lack sufficient accuracy when dealing with complex paper such as distorted, blurred, and shadowed surfaces. They also lack personalization and contextual understanding, resulting in unreliable recognition results and limited applicability to languages ​​and character sets.

Method used

The recognition model is built using deep learning and neural network technologies. Through image processing enhancement and data preprocessing, diverse text samples are collected, and customized training tools for the model are provided. Contextual information is extracted and analyzed to optimize the recognition model to adapt to complex layouts and deformed text.

Benefits of technology

It improves the recognition accuracy of scanning pens on complex paper, meets the needs of personalized users, enhances the understanding of contextual information, expands the scope of applicable languages ​​and character sets, and provides reliable recognition results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN117152768B_ABST
    Figure CN117152768B_ABST
Patent Text Reader

Abstract

本发明涉及扫读笔离线识别方法技术领域,具体为用于扫读笔的离线识别方法及系统,包括以下步骤:收集大量的书面和手写文本样本,通过图像处理增强功能进行预处理,生成文本数据集。本发明中,通过数据采集与预处理工作收集多样的样本数据,并使用图像处理技术消除纸张扭曲、模糊和阴影问题,利用深度学习和神经网络构建个性化识别模型,根据自己的写字风格或特定行业的术语进行定制化训练,整合上下文感知识别模块,提取文档结构、段落、标题等上下文信息,增强识别的准确性和语义理解,并持续扩展语言和字符集支持,满足不同地区用户的需求,最后,通过优化识别算法和模型,提高识别准确性,在处理复杂排版和变形文本时获得可靠的识别结果。
Need to check novelty before this filing date? Find Prior Art