Recognising Kazakh Sign Language with Mediapipe

The article discusses the use of the MediaPipe platform for recognising Kazakh sign language. The main focus is on the development of a system capable of real-time identification and interpretation of gestures used in Kazakh sign language. The paper describes in detail the process of setting up and training a model based on neural network architecture, as well as integration with MediaPipe to ensure high accuracy and speed. Special attention is given to data processing and annotation, as well as methods to improve recognition accuracy. As a result, an LSTM neural network capable of recognising 30 camera gestures in real time and converting them into text has been developed and tested. A small web interface was also developed for convenient interaction between the neural network and the user. Prospects for further development include expanding the set of recognisable gestures and integration with various applications to increase the level of inclusiveness.