Unlocking the Power of Al: A Real-Time Translation of Sign Language to Text

Sign language is an important means of communication for deaf and mute individuals all around the globe. It provides them with the opportunity to interact and socialize with the rest of society without feeling left out. But a large majority of the population does not know Sign Language. This paper proposes a Sign Language Recognition system based on a deep learning model to recognize American sign language alphabets (A-Z). The model uses MobileNet V1 pre-trained model to detect hand signs from the self-captured dataset. Using the Hand tracking module for data collection, the problem of lack of racial diversity was solved. The predicted labels are presented in the form of text. The complete implementation of the system is done on NVIDIA® Jetson Nano™. The model achieved a training accuracy of 99.74%. The system detects American Sign Language gestures in real time and translates them, making communication with the hearing impaired easier for everyone.