DEAFNET: Deaf-friendly Communication Network
Year: 2024
Type: conference-paper
Abstract: Hand gestures play a crucial role in sign language, facilitating non-verbal communication primarily for individuals facing hearing or speech challenges. Sign language systems, developed globally, aim to aid the communication needs of the deaf and mute community. However, existing systems often lack flexibility and cost-effectiveness. In response to these limitations, this software introduces a prototype that autonomously recognizes sign language, promoting more effective communication for individuals with hearing or speech impairments when interacting with both their peers and those without such challenges. Despite a wide social group that could profit from it, the idea of sign language recognition by technology is underutilized. Several technologies are available that can be helpful in establishing a link between this social group and the rest of the world. One of the main tools for enabling sign language users to communicate with the rest of society is understanding sign language. Computers can recognize sign language with the use of image categorization and machine learning, which can then be translated by humans. This study uses convolutional neural networks to identify sign language motions. The static sign language gestures were photographed using an RGB camera and comprise the image dataset used. The photos underwent preprocessing before being used as the input that had been cleaned. Inception v3 convolutional neural network model was used to retrain and test this dataset of sign language motions, and the findings are presented in this study. Several convolution filter inputs are processed on a single input in the model. The resulting validation accuracy was greater than 90%. The numerous attempts at sign language detection using machine learning and image depth data are also reviewed in this work. It assesses the different difficulties involved in solving the issue at hand and describes the problem's potential future.
Authors: Rajvanshi, Vanshika, Singh, Shalini, Patil, Ushmita, Prabhavathi, K
Institution: Visvesvaraya Technological University
Language: English
Subfield: Human-Computer Interaction
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W6949290643
