A Deep Learning Approach for Real-time Continuous Indian Sign Language Recognition and Translation

Indian Sign Language (ISL) serves as a vital means of interaction for people with hearing and speech impairments, yet it remains unfamiliar to most non-signers. This gap often results in communication challenges across academic, professional, and social environments. To help address these limitations, this work introduces a real-time ISL translation system that applies computer vision and deep learning techniques to interpret hand gestures and convert them into readable text or synthesized speech. The system captures user gestures through a webcam, extracts hand landmarks using MediaPipe, and classifies them with a Convolutional Neural Network (CNN). A curated dataset of frequently used ISL gestures and alphabets is used to train the model, facilitating accurate recognition and reliable performance in real-time applications.