American Sign Language Recognition Using Deep Learning

People with hearing or speaking disability require visual aids for communicating. Sign language is one such powerful means to communicate. It is the prominently used mode to express by people with hearing and speech inabilities. The research in this area has seen a rise in technology enabled gesture-based interfaces which demands a growing need for precise hand gesture identification systems. The use of deep learning-based approaches is found to be successful in various similar studies of symbol recognition and gesture identification. This work presents insights of research in context of automatic hand spelling identification with focus on exploring deep learning-based techniques. This research proposes one such strategy that applies concepts of Long Short-Term Memory and Media Pipe. It applies LSTM networks, to recognize hand gestures for sign spelling recognition and uses Media Pipe, a use instrument for identifying the movements and structural dynamics of human hands, to improve the recognition system's accuracy and resilience. The research trains and tests American sign language.