AI -Powered Sign Language Interpreter Using CNN

Abstract

Communication is a fundamental human right; however, individuals who are deaf or hearing impaired often face significant barriers when interacting with the hearing community. This paper presents a portable assistive system designed to translate hand-sign gestures into text and speech in real time, enabling more inclusive communication. The system utilizes an onboard camera combined with computer vision techniques, including MediaPipe and OpenCV, to capture and track hand movements accurately. A neural network model is developed and trained for robust key-point detection, allowing precise recognition and classification of hand gestures. Once a gesture is identified, it is instantly converted into readable text and synthesized speech, providing a dual-mode communication interface suitable for both personal and public interactions. The proposed system achieves an accuracy of approximately 90%, demonstrating reliable performance across a range of commonly used hand-sign gestures. The solution is portable, efficient, and suitable for deployment in real-world environments such as educational institutions, workplaces, and public spaces. By integrating computer vision, machine learning, and embedded systems, this work contributes to the advancement of practical and accessible assistive technologies for the deaf and hearing-impaired community.