Scholarlink

<p>Communication with deaf and hard-of-hearing people in Morocco, representing around 500,000 individuals, remains a major challenge due to the lack of effective recognition systems for sign language. This paper proposes a Moroccan sign language recognition system based on an ensemble learning approach combined with a hard voting method. We have designed a model integrating several classification algorithms in order to take advantage of their respective strengths, thus reducing classification errors through prediction fusion. By training the model on a database comprising 28 signs representing the Moroccan alphabet, we achieved a remarkable accuracy of 99%. The results demonstrate a significant improvement over traditional methods, providing a valuable resource to facilitate communication with this community. We also discuss the challenges encountered during development, such as gesture variability and data quality, as well as prospects for applying this system in real-life contexts.</p>

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

Recognition of Moroccan sign language based on a weighted ensemble learning approach

  • 期刊论文

  • 发表日期：

    2025年6月9日

查看全文
  • Meryem Cherrate,
  • My Abdelouahed Sabri,
  • Ali Yahyaouy,
  • Abdellah Aarab

摘要

Communication with deaf and hard-of-hearing people in Morocco, representing around 500,000 individuals, remains a major challenge due to the lack of effective recognition systems for sign language. This paper proposes a Moroccan sign language recognition system based on an ensemble learning approach combined with a hard voting method. We have designed a model integrating several classification algorithms in order to take advantage of their respective strengths, thus reducing classification errors through prediction fusion. By training the model on a database comprising 28 signs representing the Moroccan alphabet, we achieved a remarkable accuracy of 99%. The results demonstrate a significant improvement over traditional methods, providing a valuable resource to facilitate communication with this community. We also discuss the challenges encountered during development, such as gesture variability and data quality, as well as prospects for applying this system in real-life contexts.