Hand Sign Recognition of Indonesian Sign Language System SIBI Using Inception V3 Image Embedding and Random Forest

This paper presents a sign language recognition system for the Indonesian Sign Language System SIBI using image embeddings combined with a Random Forest classifier. A dataset comprising 5280 images across 24 classes of SIBI alphabet symbols was utilized. Image features were extracted using the Inception V3 image embedding, and classification was performed using Random Forest algorithms. Model evaluation conducted through K-Fold cross-validation demonstrated that the proposed model achieved an accuracy of 59.00%, an F1-Score of 58.80%, a precision of 58.80%, and a recall of 59.00%. While the performance indicates room for improvement, this study lays the groundwork for enhancing sign language recognition systems to support the preservation and broader adoption of SIBI in Indonesia.