This paper discusses the development of a Kazakh Sign Language recognition system and the enhancement of its effectiveness. An urgent task is the creation of a system capable of recognizing Kazakh Sign Language in real-time to facilitate communication among individuals with hearing and speech impairments. The proposed method employs a YOLOv5 convolutional neural network and the Mediapipe library, ensuring high-precision real-time gesture recognition. For the analysis and semantic processing of the recognized signs, a Long Short-Term Memory (LSTM) network is utilized. The developed system allows for the real-time analysis of users' hand movements and the generation of meaningful sentences from the recognized gestures. Furthermore, a web application based on the Django framework was developed to conveniently present the results to users. Experimental results demonstrated that the proposed system provides high accuracy and reliability in recognizing Kazakh Sign Language in real-time.