Deteksi Kesalahan Baca Tajwid Dengan Metode Deep Learning

Ramadhani, Ahmad Fatih (2026) Deteksi Kesalahan Baca Tajwid Dengan Metode Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025221191-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only Download (13MB) | Request a copy

Abstract

Kesalahan dalam membaca Al-Qur’an dapat mengubah makna ayat yang disampaikan. Perkembangan teknologi dalam bidang speech recognition memungkinkan dikembangkannya sebuah alat untuk mendeteksi kesalahan bacaan secara otomatis, sehingga dapat digunakan untuk proses belajar mandiri. Dibandingkan dengan penelitian sebelumnya yang memiliki kekurangan, yaitu: hanya mempelajari variasi hukum terbatas (Mad, Ghunnah dan Qalqalah), ketergantungan terhadap ekstraksi fitur MFCC, dan penggunaan model murni yang hanya dapat memproses pola tertentu (spasial/temporal). Tugas akhir ini mengembangkan dataset primer dan model deep learning untuk melakukan klasifikasi hukum yang lebih lengkap, khususnya berasal dari interaksi Nun Sukun/Tanwin dengan huruf lain, yaitu Idgham Bighunnah, Idgham Bilaghunnah, Ikhfa, Izhar, dan Iqlab.
Tugas Akhir ini menggunakan dataset primer sebanyak 2.404 audio dari 20 pembaca dengan interval umur 15 sampai 25 tahun, yang kemudian diproses melalui augmentasi. Dataset ini berisi rekaman simulasi kesalahan bacaan tajwid dari berbagai ayat yang diambil berdasarkan variasi huruf dan hukum. Fitur suara diekstraksi menggunakan metode Power Normalized Cepstral Coefficients (PNCC) yang lebih tahan terhadap noise dan Mel-Frequency Cepstral Coefficients (MFCC) sebagai pembanding, kemudian dilatih menggunakan model hibrida Convolutional Neural Network (CNN) dan Bidirectional Long Short-Term Memory (BiLSTM), yang kemudian dibandingkan menggunakan Bidirectional Long Short-Term Memory (BiLSTM) murni, Convolutional Neural Network (CNN) murni, dan Long Short-Term Memory (LSTM). Evaluasi dilakukan menggunakan metrik akurasi dan F1-Score. Hasil terbaik diperoleh dari kombinasi model BiLSTM dan ekstraksi fitur PNCC yang mencapai akurasi 87,34% dan F1-Score 0,8741.
====================================================================================================================================
Mistakes in reading the Quran can alter the meaning of the verses. Technological advances in speech recognition have enabled the development of a tool to automatically detect misreadings, enabling it to be used for self-learning. Compared to previous research, which had shortcomings, including only studying a limited variety of rulings (Mad, Ghunnah, and Qalqalah), relying on MFCC feature extraction, and using a pure model that can only process certain patterns (spatial/temporal), this final project develops a primary dataset and deep learning model to perform a more comprehensive classification of rulings, specifically those derived from the interaction of Nun Sukun/Tanwin with other letters, namely Idgham Bighunnah, Idgham Bilaghunnah, Ikhfa, Izhar, and Iqlab. This final project uses a primary dataset of 2,404 audio recordings from 20 readers aged 15 to 25 years, which were then processed through augmentation. This dataset contains simulated recordings of Tajweed reading errors from various verses, based on letter variations and rulings. Voice features were extracted using the Power Normalized Cepstral Coefficients (PNCC) method which is more robust to noise and Mel-Frequency Cepstral Coefficients (MFCC) as a comparison, then trained using a hybrid model of Convolutional Neural Network (CNN) and Bidirectional Long Short-Term Memory (BiLSTM), which was then compared using pure Bidirectional Long Short-Term Memory (BiLSTM), pure Convolutional Neural Network (CNN), and Long Short-Term Memory (LSTM). Evaluation was carried out using accuracy and F1-Score metrics. The best results were obtained from the combination of the BiLSTM model and PNCC feature extraction which achieved an accuracy of 87.34% and an F1-Score of 0.8741.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Tajwid, Deep Learning, CNN, BiLSTM, MFCC, PNCC.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence T Technology > T Technology (General)
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Ahmaf Fatih Ramadhani
Date Deposited:	24 Jul 2026 07:56
Last Modified:	24 Jul 2026 07:56
URI:	http://repository.its.ac.id/id/eprint/137404

Actions (login required)

View Item