ANALYSIS OF RECOGNITION ALGORITHMS AND CONVOLUTIONAL NEURAL NETWORK FOR HAND GESTURE RECOGNITION IN KAZAKH SIGN LANGUAGE

Abstract

Abstract. The ever-growing interest in machine learning and neural networks is fueled by significant advancements in computational capabilities, enabling breakthroughs in object, sound, text, and other forms of data recognition. These advancements have paved the way for a more intuitive interaction between humans and machines, making such technologies accessible to a wider audience. Recent developments in computer vision, in particular, have led to the creation of sophisticated models capable of recognizing objects in images and videos. This same technology has been effectively adapted for hand gesture recognition, enabling applications in fields like human-computer interaction, robotics, and sign language interpretation. This paper explores some of the most popular hand gesture recognition models, with a particular focus on Convolutional Neural Networks (CNNs) and Support Vector Machines (SVMs). These models differ in their methodologies, processing efficiency, and the volume of training data they require, offering various advantages and limitations depending on the application context. The core objective of this study is to provide an overview of diverse machine learning algorithms, delving deeply into their theoretical underpinnings, operational mechanisms, and comparative performance in terms of accuracy, training time, and data requirements. In addition, this work presents experimental results of sign language recognition in Kazakh, specifically using the dactyl alphabet. A detailed analysis is provided, accompanied by a comprehensive table that reports the accuracy of each recognized gesture. Real-time testing was conducted with individual hand gestures displayed in front of a camera, showcasing the effectiveness of the recognition system. Furthermore, the study incorporates an explanation of the mathematical foundations and logical structures underlying machine learning algorithms, illustrated through formulae, functional relationships, and flowcharts that depict the recognition process. By combining theoretical insights with practical experiments, this paper aims to contribute to the growing field of gesture recognition and its applications in accessible communication technologies.

Keywords: Hand gesture recognition, neural networks, algorithm, layer, CNN, SVM, YOLO.

For citation: Les N.N., Mukhanov S.B., Ipalakova M.T. ANALYSIS OF RECOGNITION ALGORITHMS AND CONVOLUTIONAL NEURAL NETWORK FOR HAND GESTURE RECOGNITION IN KAZAKH SIGN LANGUAGE//INTERNATIONAL   JOURNAL   OF   INFORMATION   AND COMMUNICA-TION TECHNOLOGIES. 2025.