Indonesian Sign Language (BISINDO) Alphabet Detection Model Using Long Short Term Memory Algorithm

Abstract

Indonesian Sign Language (BISINDO) is one of the primary communication methods for individuals with hearing disabilities in Indonesia. While communication is a fundamental human right, many non-disabled individuals still lack awareness of sign language and its significance for the deaf community, necessitating technology-based solutions to enhance communication accessibility. This study aims to develop a BISINDO alphabet detection model using deep learning with the Long Short-Term Memory (LSTM) algorithm. The dataset consists of 26 BISINDO alphabet letters, with each sample comprising 30 frames per second (fps), extracted using Mediapipe Holistic and converted into numerical arrays. This data serves as input for training the LSTM model to recognize BISINDO alphabet movement patterns. Model evaluation using accuracy metrics such as precision, recall, and F1-score demonstrates high performance, with each metric achieving 95%. These results indicate that the proposed model effectively predicts sign language gestures. This research is expected to enhance public understanding of BISINDO and contribute to the advancement of AI-based sign language translation systems.