Scholarlink

The project aims to recognize sign language using hand gestures in real accurately. Hand gesture recognition helps reduce the gap between people who are hard of hearing or are not able to speak and who can hear and speak well. The system receives sequential data representing gestures taken using a camera. In order to extract useful information from hand gestures, preprocessing techniques are utilized to increase the quality of the input data, followed by feature extraction. The recognition outcome depends on the quality of the input data, efficacy of feature extraction, structure of the recognition model, richness of the training dataset, and accuracy of the training model during real-time scenarios. Using LSTM and MediaPipe Holistics, the model archives an accuracy of around 97.4% across different dynamic signs (600 clips, 15 classes) along with static signs (1444 images, 39 classes). This study demonstrates the efficacy of the proposed system in accurately recognizing sign language gestures, thereby facilitating improved communication for individuals with hearing or speech impairments.

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

Deep Learning-Based Sign Language Detection and Prediction for Enhanced Communication

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Krisha Zalaria,
  • Nishit Dadhaniya,
  • Priyanka Pate

摘要

The project aims to recognize sign language using hand gestures in real accurately. Hand gesture recognition helps reduce the gap between people who are hard of hearing or are not able to speak and who can hear and speak well. The system receives sequential data representing gestures taken using a camera. In order to extract useful information from hand gestures, preprocessing techniques are utilized to increase the quality of the input data, followed by feature extraction. The recognition outcome depends on the quality of the input data, efficacy of feature extraction, structure of the recognition model, richness of the training dataset, and accuracy of the training model during real-time scenarios. Using LSTM and MediaPipe Holistics, the model archives an accuracy of around 97.4% across different dynamic signs (600 clips, 15 classes) along with static signs (1444 images, 39 classes). This study demonstrates the efficacy of the proposed system in accurately recognizing sign language gestures, thereby facilitating improved communication for individuals with hearing or speech impairments.