Recent Advances of Deep Learning for Sign Language Recognition

To assist the social interaction of deaf and hearing impaired people, efficient interactive communication tools is expected. With the growing research interest in action and gesture recognition in the last years, many successful applications for sign language recognition comprise new types of sensors including low-cost depth camera and advanced machine learning technologies. In this paper, we present a complete overview of deep learning based methodologies for sign language recognition. We discuss various types of such approaches designed for the recognition from viewpoints of available modalities provided by depth sensors, feature extraction and classification. In addition, we summarise the currently available datasets of sign language, including gestures of finger spelling and vocabulary words, which can be used as an assessing tool for those people who are learning sign languages. We then discuss the main current research works with particular interest on how they treat the different types of data, discussing their main features and identify opportunities and challenges for future research.