A Systematic Literature Review on the Robustness of Sign Language Recognition Methods in Low-Light Environments
AbstractThis paper presents a comprehensive review of the robustness of hand gesture recognition systems in low light intensity environments spanning the years 2018 to 2023. The primary objective is to assess the progress made during this period and identify areas requiring further attention. An extraction of 20 relevant journals from reputable online databases was conducted using selected keywords. Most of the reviewed articles delve into three crucial aspects of hand gesture recognition systems: data acquisition, data environment, and hand gesture representation. The system performance evaluation reveals that machine learning models achieve recognition accuracy between 94% and 98%, while computer vision models report accuracy within the range of 90% to 95%. The deep learning approach shows a broader accuracy range, spanning from 90% to 98%. Notably, the studies reviewed utilized datasets comprising 37 hand gestures, including 26 letters of American Sign Language (ASL) and numeric gestures ranging from 0 to 9. This paper sheds light on the current state of hand gesture recognition in low light environments and provides insights into potential opportunities for further research and development.