Dynamic Hand Gesture Recognition for Sign Language: Translating Sign Language through Gesture Patterns

Abstract

Sign language is the key to information exchange in the Deaf and Hard of Hearing Communities. Nevertheless, the communication between signers and non-signers will be constrained if interpreters must fulfil the translation process from sign language to spoken or written language. Over the past few years, the flourishing computer vision and machine learning have come out in a sprouting of highly competitive hand gesture recognition systems based on sign language translation. This article introduces a new method for the recognition of dynamic hand movements for translating sign language into text dependent on the feature extraction and distinction of the gesture patterns through Convolutional Neural Networks (CNN). The proposed algorithm traces and encodes fast hand gestures in living moments. Features are retrieved, and thus, the gestures of sign language are correctly categorized and presented in text or speech-language format. The paper summarizes the training and testing of the neural networks, the algorithm outputting promising results in the aspect of accuracy and efficiency, finding its applicability in the expanding communication of the deaf and hard of hearing community.