Rancang Bangun Aplikasi untuk Deteksi Alfabet Bahasa Isyarat Sibi menggunakan Algoritma Convolutional Neural Network (CNN)

  • Fadly Maulana Adiyasa Informatika, Universitas Persada Y.A.I, Jakarta Pusat
  • Angries Angries Informatika, Universitas Persada Y.A.I, Jakarta Pusat

Abstract

Communication is a vital necessity, however, the deaf community often faces barriers when
interacting with the general public, who typically have limited understanding of sign language.
This research aims to develop an Android-based application that translates the Indonesian Sign
Language System (SIBI) into text in real-time to facilitate such communication. The system
utilizes deep learning—specifically a Convolutional Neural Network (CNN) with a
MobileNetV2 architecture—and integrates the MediaPipe HandLandmarker framework to track
hand landmark coordinates. Classification occurs offline on the device using TensorFlow Lite to
translate the 24 static letters of the SIBI alphabet. Testing in ideal environments demonstrated
that the system achieved an average confidence level of 95.19% on smartphones and 97.50% on
tablets. Overall, the application operates responsively without memory leaks (crashes) and
robustly handles two-handed gestures. Nevertheless, feature extraction performance may
decline under extreme environmental conditions, such as strong backlighting or cluttered visual
backgrounds

Published
2026-11-03
How to Cite
Adiyasa, F. M., & Angries, A. (2026). Rancang Bangun Aplikasi untuk Deteksi Alfabet Bahasa Isyarat Sibi menggunakan Algoritma Convolutional Neural Network (CNN). IKRA-ITH Informatika : Jurnal Komputer Dan Informatika, 10(3), 207-215. Retrieved from https://journals.upi-yai.ac.id/index.php/ikraith-informatika/article/view/8015