Scholarlink

<p>Sign language is the primary mode of communication for individuals with speech and hearing impairments, relying heavily on hand gestures, finger shapes, and movements. Recognizing dynamic sign language words remains a significant challenge due to limited datasets and complex features. This study proposes a novel approach for classifying 24 dynamic American Sign Language (ASL) words using a Leap Motion Controller (LMC), which provides skeletal hand and finger data. Two custom datasets were created: one for 13 single-hand words and another for 11 double-hand words. A custom feature set was extracted, and a stacked Long Short-Term Memory (LSTM) model was employed for classification. The proposed method demonstrated high accuracy across multiple train-test-validation splits, showcasing its effectiveness in dynamic sign language recognition.</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

Deep learning based dynamic sign language translation system

  • 期刊论文

  • 发表日期：

    2025年4月23日

查看全文
  • Batool A. Sahm,
  • Hassanin Al-Fahaam,
  • Abbas A. Jasim

摘要

Sign language is the primary mode of communication for individuals with speech and hearing impairments, relying heavily on hand gestures, finger shapes, and movements. Recognizing dynamic sign language words remains a significant challenge due to limited datasets and complex features. This study proposes a novel approach for classifying 24 dynamic American Sign Language (ASL) words using a Leap Motion Controller (LMC), which provides skeletal hand and finger data. Two custom datasets were created: one for 13 single-hand words and another for 11 double-hand words. A custom feature set was extracted, and a stacked Long Short-Term Memory (LSTM) model was employed for classification. The proposed method demonstrated high accuracy across multiple train-test-validation splits, showcasing its effectiveness in dynamic sign language recognition.