Abstract
This systematic review explores the importance of the Kazakh Sign Language (KSL) as the main method of communication for deaf and hard-of-hearing people in Kazakhstan. Despite its critical role, the automatic translation of KSL into spoken language remains underdeveloped. The purpose of this systematic review is to examine recent progress in dynamic gesture recognition technologies aimed at translating KSL into speech. This study emphasizes the use of machine learning models, particularly Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, to identify KSL gestures and assess their efficiency in real-time translation systems. The research approach involves reviewing existing studies on these models, highlighting their strengths and weaknesses, and evaluating their effectiveness in KSL gesture recognition. The results obtained demonstrate that while CNN and LSTM models show significant promise for real-time translation, challenges such as limited datasets and model generalization persist. The work crucially contributes to the enhancement of accessibility and promotion of social inclusion by offering comprehensive analysis and providing recommendations for future research directions aimed at developing efficient, real-time KSL-to-speech translation systems.