Scholarlink

The scientific community is still far from having a practical solution for effectively and reliably comprehending handwritten Arabic characters even with the extensive use of Deep Learning models in fields like computer vision, automated recognition of objects, and natural language processing. The lack of published research that expressly addresses the complex nature of this linguistic issue as compared to other languages contributes to this challenge. Arabic handwritten character identification is challenging, especially when accurately recognizing characters from visual sources and understanding their meaning. The main source of these difficulties is the fundamental variety seen in people’s handwriting, which is further complicated by alterations in writing styles throughout time. To address these issues, this study provides a hybrid model that combines the capabilities of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The suggested approach consists of two main stages: first, a CNN is employed to extract detailed information from handwritten Arabic characters. These attributes are then input into an LSTM network, which captures the script’s sequential dependencies and contextual information. The enhanced performance indicates the efficacy of combining feature extraction and sequence modeling to overcome traditional recognition limitations. The suggested model consistently produced a noteworthy character detection accuracy rate of around 96%.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Evaluating CNN and Hybrid CNN-LSTM Models for Arabic Handwritten Character Recognition

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Othmane Farhaoui,
  • Mohamed Rida Fethi,
  • Ali Omari Alaoui,
  • Imad Zeroual,
  • Ahmad El Allaoui

摘要

The scientific community is still far from having a practical solution for effectively and reliably comprehending handwritten Arabic characters even with the extensive use of Deep Learning models in fields like computer vision, automated recognition of objects, and natural language processing. The lack of published research that expressly addresses the complex nature of this linguistic issue as compared to other languages contributes to this challenge. Arabic handwritten character identification is challenging, especially when accurately recognizing characters from visual sources and understanding their meaning. The main source of these difficulties is the fundamental variety seen in people’s handwriting, which is further complicated by alterations in writing styles throughout time. To address these issues, this study provides a hybrid model that combines the capabilities of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks. The suggested approach consists of two main stages: first, a CNN is employed to extract detailed information from handwritten Arabic characters. These attributes are then input into an LSTM network, which captures the script’s sequential dependencies and contextual information. The enhanced performance indicates the efficacy of combining feature extraction and sequence modeling to overcome traditional recognition limitations. The suggested model consistently produced a noteworthy character detection accuracy rate of around 96%.