Klasifikasi Hukum Tajwid Otomatis Pada Resitasi Al-Qur’an Menggunakan Deep Learning Di Pp Syaichona Cholil Putri Balikpapan

Abstract

In Islamic education traditions, the accuracy of reading the Qur’an in accordance with tajweed rules is a crucial aspect of the quality of students’ recitation. However, the limited number of educators and the need for repeated practice pose challenges in the learning process. This study aims to develop an automatic tajweed rule classification system based on deep learning as an evaluation tool for students’ recitation at PP Syaichona Cholil Putri Balikpapan. This research uses an applied quantitative-experimental approach with a system development design. Data were obtained through student voice recordings, tajweed rule annotation by expert teachers, signal processing, and feature extraction using mel-spectrogram/MFCC. The CRNN model was tested using cross-validation and field trials, accompanied by questionnaires and interviews to assess its usefulness. The results show that the CRNN model is capable of classifying priority tajweed rules with an accuracy of 92–95% on internal test data. Classroom implementation received positive responses as it effectively detects minor errors quickly and systematically, thereby accelerating correction and monitoring students’ progress. This study asserts that artificial intelligence is effective as a learning companion while still positioning the teacher as the main validator. This system contributes to improving the quality of tajweed learning in Islamic boarding schools and enriches deep learning research based on local datasets. In the future, it is necessary to increase dataset coverage, vary recording environments, and integrate web/smartphone platforms to broaden the model's application.