A systematic review of sign language classification system for real-time application

Sign language (SL) has been a crucial form of communication for the deaf and hard of hearing (DHH) community for the longest time. SL allows DHH individuals to express their thoughts and ideas without verbalizing. However, there are significant challenges to ensure its accessibility to a broader society due to the lack of understanding SL. This creates a communication barrier among the community widely. With the current advancement of recognition systems that use deep learning (DL), an automatic SL recognition system can be developed to bridge the communication gap. This study performs a systematic literature review (SLR) that explores the different approaches for SL classification system, common challenges in SL classification and the hardware implementation for SL classification. Through this study, 737 papers were selected and have been narrowed down to 85 papers that have been thoroughly viewed based on regions of interest of DL vision-based SL classifications system. Additionally, SL classification system papers published between 2020 to 2024 have been studied and analyzed. The presented findings prove that DL vision-based system is the suitable approach for SL classification while overcoming the main challenge addressed which is environmental factors. Finally, the SLR also proves FieldProgrammable Gate Array (FPGA) as the viable hardware for DL vision-based SL classification system.