Translasi Bahasa Isyarat Berstandar Bisindo ke Bahasa Verbal Indonesia Menggunakan Algoritma YOLO dan LSTM

Abstract

Communication is a crucial aspect of social life. Language serves as a means of exchanging information. Verbal language, both spoken and written, is commonly used for communication. However, there are individuals who are unable to use verbal language effectively due to their limitations. Generally, they resort to using sign language, a nonverbal form of communication. Sign language is not commonly learned and used by those who do not have limitations in verbal communication. As a result, communication between those who use verbal and nonverbal language faces challenges in conveying information. To address this issue, this research proposes the implementation of an object detection system using YOLO (You Only Look Once) and a text normalization system with a lexical approach using LSTM (Long Short-Term Memory) algorithm. The aim of this study is to translate sign language gestures in BISINDO (Indonesian Sign Language) standard into improved textual representations. The results of applying YOLO in detecting sign language gestures show a highest confidence box of 0.95, and the implementation of LSTM in normalizing the output of sign language gesture detection achieves a highest accuracy of 0.8