Deteksi Kesalahan Bacaan Tajwid pada Audio Al-Qur'an menggunakan Arsitektur Transformer

Ni'am, Muhammad Choirun (2026) Deteksi Kesalahan Bacaan Tajwid pada Audio Al-Qur'an menggunakan Arsitektur Transformer. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221203-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (8MB) | Request a copy

Abstract

Membaca Al-Qur’an dengan tartil dan sesuai kaidah ilmu tajwid merupakan kewajiban bagi setiap muslim. Kesalahan yang sering dijumpai terjadi dalam tilawah adalah kesalahan pelafalan sifat huruf, seperti pantulan pada hukum Qalqalah, serta ketidasesuaian durasi pada hukum Mad. Metode pembelajaran konvensional dengan evaluasi dari pengajar manusia seringkali terkendala oleh keterbatasan ketersediaan pengajar dan penilaian yang tidak konsisten. Penelitian terdahulu yang menggunakan fitur manual seperti MFCC dengan model CNN atau RNN memiliki keterbatasan dalam menangkap nuansa akustik halus pada sinyal audio. Tugas Akhir ini mengembangkan sistem deteksi kesalahan tajwid otomatis menggunakan arsitektur Transformer berbasis model Wav2Vec 2.0 Large XLSR-53 Arabic. Dataset dikumpulkan dari 20 responden yang mencakup 11 kelas hukum Qalqalah dan Mad (Thabi’i, Wajib Muttasil, dan Lazim). Strategi transfer learning diterapkan dengan metode ablation study dengan menggunakan konfigurasi lima freeze layer (0, 6, 12, 18,dan 24). Hasil eksperimen menunjukkan bahwa konfigurasi dengan freeze layer 6 merupakan konfigurasi paling optimal dengan akurasi mencapai 94,06%, unggul 12,78% dibanding model CRNN yang mencapai akurasi 81,28%. Uji signifikansi McNemar memberikan hasil signifikan (p<0,05). Sistem diimplementasikan dalam aplikasi untuk memudahkan pengguna mendeteksi kesalahan tajwid melalui audio secara langsung.
======================================================================================================================================
Reciting the Al-Qur’an with tartil and adhering to Tajwid rules is important for every Muslim. Common errors in recitation include inaccuracies in articulating letter characteristics, such as the bouncing sound in Qalqalah, and imprecise duration in Mad rules. Conventional learning methods relying on subjective evaluation by human teachers are often hindered by limited availability and inconsistency in assessment. Previous studies utilizing handcrafted features like MFCC with CNN or RNN models have limitations in capturing subtle acoustic nuances in audio signals. This Final Project develops an automated Tajwid error detection system using a Transformer architecture based on the Wav2Vec 2.0 Large XLSR-53 Arabic model. The dataset collected from 20 respondents covering 11 classes of Qalqalah and Mad (Thabi'i, Wajib Muttasil, and Lazim) rules. A transfer learning strategy was applied with an ablation study on five freeze layer configurations (0, 6, 12, 18, and 24) to determine the optimal configuration. Experimental results show that the freeze layer = 6 configuration achieves the best performance with an accuracy of 94.06%. The optimized CRNN model achieves an accuracy of 81.28%, resulting in a 12.78% advantage for Wav2Vec 2.0. McNemar's statistical significance test confirms that the performance difference between both models is statistically significant (p < 0.05). The system is implemented as an application enabling users to detect Tajwid errors through audio file uploads or direct microphone recording.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Automatic Speech Recognition, Deep Learning, Deteksi Kesalahan Tajwid, Mad, Qalqalah, Transformer, Wav2Vec 2.0, Automatic Speech Recognition, Deep Learning, Mad, Qalqalah, Tajwid Error Detection, Transformer, Wav2Vec 2.0
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.9.U83 Graphical user interfaces. User interfaces (Computer systems)--Design.
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Muhammad Choirun Niam
Date Deposited:	25 Jul 2026 09:21
Last Modified:	25 Jul 2026 09:21
URI:	http://repository.its.ac.id/id/eprint/138152

Actions (login required)

View Item