Scholarlink

In the era of Industry 4.0, characterized by cutting-edge automation and digital technology, improving communication for individuals with speech impairments poses a critical challenge. This research focuses on overcoming communication barriers for the hearing-impaired population in India by developing a machine learning- based system to recognize Devnagari Sign Language (DSL) gestures. The solution aims to provide a highly accurate system to facilitating communication for those reliant on sign language. The paper reviews existing literature on various kinds of systems for detecting and interpreting sign language and discusses different approaches, including sensor-based and vision-based systems. The proposed system created a dataset of 47 alphabets of Devnagari Sign Language and applied different image augmentation techniques such as Horizontal flip, Blurring Techniques, Color Augmentations, Shift-Scale- Rotate and Hue-Saturation-Value Adjustment. This study evaluates the efficiency of a proposed system using precision, recall, F1-score, and support metrics. The proposed model attained an accuracy of 97.35% for 47 Devnagari alphabets by using LSTM based RNN.

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

Devnagari Sign Language Recognition with LSTM-Based RNN

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Deepali R. Naglot,
  • Deepa S. Deshpande

摘要

In the era of Industry 4.0, characterized by cutting-edge automation and digital technology, improving communication for individuals with speech impairments poses a critical challenge. This research focuses on overcoming communication barriers for the hearing-impaired population in India by developing a machine learning- based system to recognize Devnagari Sign Language (DSL) gestures. The solution aims to provide a highly accurate system to facilitating communication for those reliant on sign language. The paper reviews existing literature on various kinds of systems for detecting and interpreting sign language and discusses different approaches, including sensor-based and vision-based systems. The proposed system created a dataset of 47 alphabets of Devnagari Sign Language and applied different image augmentation techniques such as Horizontal flip, Blurring Techniques, Color Augmentations, Shift-Scale- Rotate and Hue-Saturation-Value Adjustment. This study evaluates the efficiency of a proposed system using precision, recall, F1-score, and support metrics. The proposed model attained an accuracy of 97.35% for 47 Devnagari alphabets by using LSTM based RNN.