Arizki, Dhaffa (2026) Aplikasi Penerjemah Bahasa Isyarat Indonesia (BISINDO) Berbasis Deep Learning Pada Aplikasi Mobile. S1 Teknik Informatika thesis, STMIK Widya Cipta Dharma.
![[img]](https://repository.wicida.ac.id/style/images/fileicons/text.png) |
Text
2243050-S1-Jurnal.pdf Download (426kB) |
|
Text
2243050-S1-Teknik Informatika.pdf Restricted to Repository staff only Download (2MB) | Request a copy |
Abstract
Kesenjangan komunikasi antara penyandang tunarungu dan masyarakat umum masih menjadi permasalahan yang cukup signifikan di Indonesia, terutama karena rendahnya pemahaman masyarakat terhadap Bahasa Isyarat Indonesia (BISINDO). Media penerjemah konvensional seperti buku panduan isyarat maupun jasa penerjemah manusia bersifat terbatas, tidak selalu tersedia, dan kurang praktis untuk digunakan dalam interaksi sehari-hari. Oleh karena itu, penelitian ini bertujuan untuk merancang dan membangun aplikasi penerjemah BISINDO berbasis deep learning pada perangkat mobile yang mampu mengenali gesture tangan alfabet A–Z secara real-time melalui kamera smartphone. Metode pengembangan sistem yang digunakan dalam penelitian ini adalah metode Prototyping yang terdiri dari tahapan Communication, Quick Plan, Modeling Quick Design, Construction of Prototype, dan Deployment, Delivery and Feedback. Sistem dibangun menggunakan metode deep learning Convolutional Neural Network (CNN) dengan arsitektur MobileNetV2 yang dioptimalkan menggunakan TensorFlow Lite (TFLite) agar dapat berjalan secara efisien pada perangkat mobile, serta diintegrasikan dengan MediaPipe Hand Landmarker untuk mendeteksi tangan pengguna sebelum proses klasifikasi dilakukan. Selain fitur penerjemah real-time, aplikasi juga dilengkapi dengan fitur Belajar BISINDO berupa video animasi gerakan alfabet yang dapat diakses secara offline sebagai media edukasi tambahan. Hasil penelitian menunjukkan bahwa aplikasi BISINDO Translator berhasil mengenali dan menerjemahkan gesture tangan alfabet A–Z secara real-time dengan baik. Hasil pelatihan model menunjukkan validation accuracy sebesar 99,96% serta nilai accuracy pada Classification Report sebesar 1,00 (100%). Berdasarkan hasil pengujian Black Box terhadap 18 skenario, seluruh fitur aplikasi berjalan sesuai dengan rancangan yang telah ditetapkan. Dengan demikian, aplikasi yang dikembangkan dapat menjadi media bantu komunikasi yang efektif antara penyandang tunarungu dan masyarakat umum. ============================================================ The communication gap between deaf individuals and the general public remains a significant issue in Indonesia, largely due to the public's limited understanding of Indonesian Sign Language (BISINDO). Conventional translation media, such as sign language guidebooks or human interpreter services, are limited, not always available, and impractical for everyday interaction. Therefore, this study aims to design and develop a deep learning-based BISINDO translator application on a mobile device capable of recognizing A–Z alphabet hand gestures in real time using a smartphone camera. The system development method used in this research is the Prototyping method, which consists of the Communication, Quick Plan, Modeling Quick Design, Construction of Prototype, and Deployment, Delivery and Feedback stages. The system was built using a deep learning approach with a Convolutional Neural Network (CNN) and MobileNetV2 architecture, optimized with TensorFlow Lite (TFLite) to run efficiently on mobile devices, and integrated with MediaPipe Hand Landmarker to detect the user's hand prior to the classification process. In addition to the real-time translation feature, the application is also equipped with a BISINDO Learning feature consisting of animated videos of alphabet gestures that can be accessed offline as a supplementary educational medium. The results indicate that the BISINDO Translator application successfully recognizes and translates A–Z alphabet hand gestures in real time. The model training results show a validation accuracy of 99.96% and a Classification Report accuracy of 1.00 (100%). Based on Black Box testing on 18 scenarios, all application features worked according to the established design. Therefore, the developed application can serve as an effective communication aid between deaf individuals and the general public.
| Item Type: | Thesis (S1 Teknik Informatika) |
|---|---|
| Additional Information: | Pembimbing 1 : Siti Lailiyah, S.Kom., M.Kom Pembimbing 2 : Dr. Syamsuddin Mallala, M.Pd |
| Uncontrolled Keywords: | BISINDO, Deep Learning, MobileNetV2, TensorFlow Lite, MediaPipe Hand Landmarker. |
| Subjects: | Q Science > QA Mathematics > QA76 Computer software |
| Divisions: | Teknik Informatika |
| Depositing User: | Mr Dhaffa Arizki |
| Date Deposited: | 06 Aug 2026 03:48 |
| Last Modified: | 06 Aug 2026 03:48 |
| URI: | http://repository.wicida.ac.id/id/eprint/6370 |
Actions (login required)
 |
View Item |
