Sign Language Recognition: A Comprehensive Review Of Methods, Datasets, And Challenges

Abstract

Sign Language Recognition (SLR) Is An Active Research Area That Involves The Automatic Translation Of Sign Languages Into Text Or Speech Using Computational Techniques, Thereby Facilitating Communication For Deaf And Hard-of-hearing Individuals. Over The Past Two Decades, SLR Has Evolved From Traditional Handcrafted Feature-based Approaches To Modern Deep Learning-driven Methods Due To Significant Advances In Computer Vision. This Survey Reviews Approximately 15–25 Representative Studies And Categorizes Existing SLR Approaches Into Isolated And Continuous Recognition Tasks. Widely Used Datasets, Feature Extraction Techniques, Learning Models, And Evaluation Metrics Are Discussed. Classical Methods Such As Hidden Markov Models And Support Vector Machines Are Reviewed Alongside Deep Neural Architectures Including Convolutional Neural Networks, Recurrent Neural Networks, And Transformer-based Models. Key Challenges Include Data Sparsity, Signer Variability, And Difficulties In Modeling Non-manual Cues. Finally, Promising Future Research Directions Such As Multimodal Learning And Low-resource Sign Language Recognition Are Highlighted.

Keywords