Survey of Deep Learning-Based Sign Language Recognition Systems for Accurate Gesture Classification and Human–Computer Interaction

Abstract

Sign Language Recognition (SLR) is becoming an area of paramount importance in the fields of Artificial Intelligence, Computer Vision and Human Computer Interaction (HCI) because of the communication gaps it can fill among hearing impaired and the hearing community. Recently, the recognition of complex human gestures in SL has progressed significantly, with deep learning being used to automatically and accurately extract features and learn from complex SL gestures. It is a review paper that covers a comprehensive survey of deep learning based SLR systems, architectures including Convolutional Neural Networks (CNNs), Recurrent Neural Networks (RNNs), Long Short-Term Memory (LSTM) networks, Gated Recurrent Units (GRUs), Transformers, and hybrid models. In this study, the recent advances in gesture classification, multimodal gesture recognition frameworks, attention mechanisms, transfer learning methods and lightweight deployment strategies are analyzed. Moreover, the review discusses the emerging applications of multimodal sensing technologies such as RGB cameras, wearable sensors, skeletal tracking, radar systems, and facial expression analysis to boost recognition robustness and real-time performance. Other applications are also discussed such as assistive communication, healthcare, education, virtual reality, smart devices and human–robot interaction. Lastly, some of the current challenges, such as signer variation, multilingual recognition, continuous sign language translation and real world deployment of such systems are defined. The results suggest that great progress has been made with deep learning based SLR, and that it is a promising approach to enable the creation of intelligent, accessible and inclusive communication technologies.