Scholarlink

In today’s communication-centric world, there is a need to bridge the gap of connectivity and inclusivity for people with hearing and speech impairments. For such individuals, communication often poses a challenge, due to which the need of a sign language interpretation system becomes significant. India, being a culturally rich nation with linguistic diversity, a unified communication medium can be offered through Indian Sign Language. This paper aims to propose a solution to overcome this communication barrier by constructing an ISL interpretation system using transfer learning with VGG16, which is a Convolutional Neural Network model with 16 weight layers. The dataset used for validation and evaluation has been created manually for 26 English letters of the alphabet by image augmentation. There are a total of 16,900 images that gave 97.87% evaluation accuracy and 99.85% testing accuracy, upon training.

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

Design of Indian Sign Language Interpretation System Using Transfer Learning VGG16 Model

  • 会议论文

  • 发表日期：

    2026年

查看全文
  • Muskaan Anand,
  • Dhwani Arora,
  • Satwinder Kaur,
  • Shruti Arora,
  • Garima Joshi,
  • Naveen Aggarwal

摘要

In today’s communication-centric world, there is a need to bridge the gap of connectivity and inclusivity for people with hearing and speech impairments. For such individuals, communication often poses a challenge, due to which the need of a sign language interpretation system becomes significant. India, being a culturally rich nation with linguistic diversity, a unified communication medium can be offered through Indian Sign Language. This paper aims to propose a solution to overcome this communication barrier by constructing an ISL interpretation system using transfer learning with VGG16, which is a Convolutional Neural Network model with 16 weight layers. The dataset used for validation and evaluation has been created manually for 26 English letters of the alphabet by image augmentation. There are a total of 16,900 images that gave 97.87% evaluation accuracy and 99.85% testing accuracy, upon training.