A Real Time Conversion Model for Hand Gestures to Textual Content

Abstract

Sign language is a form of communication that uses hand movements and gestures to convey meaning to deaf and mute individuals. We attempted to create a real-time finger spelling system using a convolutional neural network based on American Sign Language (ASL). The paper presents the recognition of 26 alphabet hand gestures in ASL. The system has several modules, including pre-processing, training, and testing, and achieved an accuracy of 95.8% in extracting, processing, training, and testing the model, as well as converting ASL into text. In this model, we utilized deep learning, OpenCV, and TensorFlow to identify hand gestures and found that our dataset yielded improved recognition results.