Pengenalan Bahasa Isyarat Indonesia (BISINDO) Dinamis Menggunakan Mediapipe dengan Algoritma Temporal Convolutional Network (TCN)

Abstract

Communication is a fundamental human need, but for the deaf-mute community, it poses challenges due to limitations in speaking and hearing. Indonesian Sign Language (BISINDO) is the primary communication tool for the deaf-mute community, yet it is not widely understood by the general public. This study aims to develop a dynamic BISINDO recognition system using MediaPipe and the Temporal Convolutional Network (TCN) algorithm, which is expected to assist in translating BISINDO sign language into Indonesian in the form of text and voice. By employing literature review, problem analysis, system design, implementation, and testing methods, this study demonstrates that the developed system is capable of recognizing sign movements with an accuracy of 90.25%, with high precision, recall, and f1-score for each data label. The system is effective in reducing data pre-processing complexity and accurately recognizing movement patterns. In conclusion, this study successfully designed and implemented a reliable BISINDO recognition system, with suggestions for further development, including the use of larger datasets, integration with real-time technology, and the development of mobile applications to enhance accessibility.