Scholarlink

Current sign language translation systems often struggle with accuracy, real-time processing, and multilingual capabilities, presenting significant communication barriers between the hearing and deaf communities. This paper proposes an improvement to the real-time sign language recognition and translation system, the system which combines TensorFlow’s MoveNet with ResNet, DenseNet, EfficientNet, and InceptionNetV3 ensemble. This design aims to improve the communication gap between the deaf and hearing impaired by relaying sign language gesture recognition and synthesizing it to spoken language. It accepts multiple sources and has simultaneous translation capabilities in and between different languages . All results of the experiments show the effectiveness of the system achieving accuracy of the system as 98.5% percent, Precision 97.8%, Recall 98.2% F1 score of 98.0%. It does not only improve the use by deaf individuals but also improves communication across language boundaries. The system has major prospects for future optimisation of computation and the addition of more languages to its portfolio.

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

Real-Time Sign Language Recognition and Multilingual Translation Using Ensemble Deep Neural Networks and TensorFlow’s MoveNet

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Aarthy Reddy Sontam,
  • Deepika Boyina,
  • Minu Susan Jacob

摘要

Current sign language translation systems often struggle with accuracy, real-time processing, and multilingual capabilities, presenting significant communication barriers between the hearing and deaf communities. This paper proposes an improvement to the real-time sign language recognition and translation system, the system which combines TensorFlow’s MoveNet with ResNet, DenseNet, EfficientNet, and InceptionNetV3 ensemble. This design aims to improve the communication gap between the deaf and hearing impaired by relaying sign language gesture recognition and synthesizing it to spoken language. It accepts multiple sources and has simultaneous translation capabilities in and between different languages . All results of the experiments show the effectiveness of the system achieving accuracy of the system as 98.5% percent, Precision 97.8%, Recall 98.2% F1 score of 98.0%. It does not only improve the use by deaf individuals but also improves communication across language boundaries. The system has major prospects for future optimisation of computation and the addition of more languages to its portfolio.