Implementation of Machine Learning Model to Detect Sign Language Movement in SIBI Learning Media

Abstract

This research focuses on the development of a web-based Indonesian Sign Language System (SIBI) learning application with motion detection to improve the precision of sign language practice. Despite the government's introduction of SIBI as an official system, existing platforms lack tools to validate the accuracy of hand movements. Using the Design Sprint methodology—comprising Understand, Define, Sketch, Decide, Prototype, and Validate phases—this study employs Microsoft Azure Machine Learning to create a motion detection model capable of recognizing SIBI gestures. The application offers an interactive learning experience, allowing users to practice and receive real-time feedback on their accuracy. Initial trials demonstrated high prediction accuracy, achieving 99.82% on public datasets and 96.4% on private datasets. Beta testing revealed an 86% satisfaction rate among users, indicating the application’s effectiveness in enhancing the learning process. By providing accessibility through standard web browsers and incorporating advanced motion detection, this application contributes to inclusivity, facilitating broader public understanding and interest in learning sign language.