Pengembangan Sistem Klasifikasi Gestur Bahasa Isyarat Berbasis Gamifikasi Adaptif Menggunakan SVM Dengan Mekanisme Adaptive Gesture Difficulty Leveling (AGDL)

This study aims to develop an interactive, web-browser-based Sistem Isyarat Bahasa Indonesia (SIBI) gesture classification system to facilitate inclusive learning for the Deaf community. The methodology integrates MediaPipe landmark extraction and a Support Vector Machine (SVM) algorithm executed entirely via client-side inference using the ONNX format, complemented by an Adaptive Gesture Difficulty Leveling (AGDL) mechanism. Evaluation results demonstrate that the SVM model achieves an accuracy of 98.92% across 36 gesture classes, while yielding a System Usability Scale (SUS) score of 86.0 (Best Imaginable). Furthermore, the implementation of a client-side architecture successfully eliminates server latency and ensures user data privacy. Concurrently, the deployment of adaptive gamification elements dynamically maintains user engagement within the flow zone. The broader impact of this innovation lies in providing a scalable, secure, and widely adoptable educational technology (EdTech) solution that bridges communication gaps without requiring high-cost cloud infrastructure.