<![CDATA[This research presents the development of SiPuTiH (Handwritten Hijaiyah Character Recognition System) using the Convolutional Neural Network (CNN) algorithm to address the challenges of handwriting variability in Arabic scripts. The methodology includes dataset acquisition and preprocessing, CNN architecture design, model training, and performance evaluation. The dataset consists of 1,680 handwritten images representing 30 Hijaiyah characters, divided into 80% training and 20% testing data. The proposed CNN architecture employs four convolutional and pooling layers with a total of 6.8 million trainable parameters. Experimental results show that SiPuTiH achieved a 99.7% accuracy rate in recognizing Hijaiyah characters, with only one misclassification between ‘ta’ (ت) and ‘tsa’ (ث) due to morphological similarity. The trained model was implemented in an interactive Streamlit-based application that includes learning modules, quizzes, and real-time handwriting prediction. SiPuTiH demonstrates high reliability not only as a handwriting recognition system but also as an engaging educational platform for learning Arabic letters. This study confirms the effectiveness of CNNs in handling the morphological complexity of Hijaiyah characters and contributes to the development of intelligent educational tools. Future work may explore larger datasets, transfer learning architectures, and contextual (word-level) recognition to enhance system scalability and performance.]]>
Copyrights © 2025
