American Sign Language Static Gesture Recognition using Deep Learning and Computer Vision
Sai Nikhilesh Reddy Karna, Jai Surya Kode, Suneel Nadipalli, Sudha Yadav
Abstract
Specially-abled people (speech and hearing impaired) rely on hand-gestures for communication on a daily basis. Majority of the people aren’t aware of the universally accepted hand-gestures alphabet, making communication difficult between the two groups of people. In an attempt to fill this void, this research work proposes a real time hand-gesture based recognition system based on the American Sign Language (ASL) dataset and capturing data through a BGR webcam and processing it using Computer Vision (OpenCV). The 29 static gestures (the alphabet) from the official, standard ASL dataset were trained with the help of Vision Transformer Model (ViT). The model showed an accuracy rate of 99.99% after being trained with 87,000 RGB samples for 1 epoch (2719 batches of 32 images each).