Convolutional Neural Network -Based Indonesian Sign Language (BISINDO) Recognition for Enhanced Communication

Sign language recognition is a crucial step in enhancing the engagement of deaf community. The system proposed in this paper focuses on developing a real-time recognition system for a single BISINDO hand gesture and converting it into the corresponding letter using a Convolutional Neural Network (CNN) model. The model is built with three convolutional layers, followed by a pooling layer and two fully connected layers for classification. The hand pose dataset is sourced from the Kaggle platform, incorporating various gesture variations for all letters to improve the model's accuracy and generalization capability. The tangible result is a prototype in which a limited conversion is resulted from a single hand gesture into a single letter. Our results demonstrate that CNN model approach can achieve an accuracy rate of 86.06% on testing data. This project is expected to contribute in facilitating communication between the hearing and deaf communities while opening opportunities for further development, such as translating into complete sentences and real-time speech conversion.