Smart Vision for Silent Voices: A Review of Sign Language Recognition Technologies

The identification of sign language has become a leading area of study with the aim of enhancing the interaction between society and hearing-impaired individuals. This paper presents a critical overview of thirteen state-of-the-art research papers that employ deep learning, computer vision, and artificial intelligence frameworks for sign language identification and translation. The research studies investigate diverse methods, such as Convolutional Neural Networks (CNN), Long Short-Term Memory Networks (LSTM), and Mediapipe frameworks for static and dynamic sign processing. Data enhancement, gesture tokenization, and hybrid artificial intelligence models are also considered. This review critically contrasts the strength and weakness of each research study, such as their highlighting contribution, namely, recognition accuracy, model optimization, and real-time reaction. The limitations considered are the lack of sufficient data, high computational cost, and limited vocabulary recognition ability. With the collation of the outcomes of these research studies, this paper considers future research directions such as the need for large datasets, multilinguality, and enhanced real-time performance to enable the development of effective and accessible sign language translation systems.