Application of Deep Learning Techniques on Sign Language Recognition—A Survey

Abstract

Sign language recognition systems are developed to facilitate communication between signers and non-signers. Recent field of research is intended to focus on effectively recognizing signs under computing power constraints. The work primarily includes recognizing sign languages using discrete cosine transforms, principal component analysis, and hidden Markov models. Researchers have used a wide variety of machine learning and deep learning techniques such as artificial neural network, convolutional neural network, minimum distance classifier, three-dimensional residual convolutional neural networks, bidirectional long short-term memory networks, ‘CaffeNet’ convolutional neural network, and so on. Some researchers have used hand trajectories, depth-sensing cameras, etc., to detect the motion. This paper reviews the literature that has been carried out to recognize the most widely used sign languages like Indian sign language, American sign language, Persian sign language, etc., using machine learning and deep learning techniques. This paper draws similarities and differences between various sign languages and their algorithms to infer which techniques are best suited for Indian sign language recognition.