Dynamic Sign Language Translator

Sign language is a visual-based communication that utilizes hand movements and variations of hand shape to express speech and is the chief communication instrument for people with listening and speech impairments. The proposed Automated sign language translation work can aid such people who require sign language to communicate, have intention of enhanced usage of sign language. The work also can help people them to learn sign language, and present a convenient means of education, profession and life for people. Hand locating and sign language recognition methods can generally be divided into traditional methods and deep learning methods. With the glorious accomplishments of deep learning in the domain of computer vision, demonstrated that deep learning has several benefits, such as feature extraction and reliable modeling capability. Consequently, this paper studies various prepossessing and training techniques for automated sign language recognition, including Mediapipe for coordinates extraction, labeling each frame to replicate Ensemble techniques, and vertically stacking frames to test One-Shot techniques. The models were tested with a data-set that was prepared of 15 phrases with about 20 videos for each phrase to achieve 90% + accuracy.