Sign Language Translation Via Computer Vision Technique

Communication barriers between deaf and hard-of-hearing individuals and the general population remain a significant challenge, particularly in regions where Indian Sign Language (ISL) is not widely understood. This paper presents a real-time Indian Sign Language translation system based on computer vision and deep learning techniques. The proposed system captures live hand, face, and body movements using a standard webcam and processes these visual cues to recognize commonly used ISL gestures. MediaPipe Holistic is employed for keypoint extraction, and a Long Short-Term Memory (LSTM) network is used to classify dynamic gesture sequences. Recognized gestures are translated into readable text or audible speech to facilitate seamless interaction with hearing individuals. The system is designed to be lightweight and efficient, ensuring real-time performance on affordable consumer hardware while addressing challenges such as lighting variations, background noise, and gesture diversity. The proposed solution aims to promote accessibility, inclusivity, and independent communication in everyday environments.