Pengenalan Maqam pada Tilawah (Seni Baca Al-Qur’an) dengan Gaya Mujawwad Menggunakan Deep Learning

Sholeh, Arief Badrus (2024) Pengenalan Maqam pada Tilawah (Seni Baca Al-Qur’an) dengan Gaya Mujawwad Menggunakan Deep Learning. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 5025201228-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 October 2026. Download (7MB) | Request a copy

Abstract

Penelitian ini bertujuan untuk mengembangkan model pengenalan maqam pada tilawah mujawwad serta aplikasi berbasis Android sebagai antarmuka inferensi model. Pengenalan maqam pada tilawah mujawwad penting bagi mereka yang ingin mendalami seni tilawah ini, namun sering menghadapi kendala dalam memahami perbedaan karakteristik melodi, ritme, dan perubahan nada yang unik pada setiap jenis maqam. Selain itu, mempelajari tilawah membutuhkan bimbingan khusus dari guru Al-Qur’an. Tanpa kehadiran guru, murid akan menghadapi kesulitan dalam penerapan maqam yang mungkin tidak tepat. Sehingga, diperlukan pendekatan lain untuk mengatasi permasalahan tersebut, yaitu melalui pemanfaatan Deep Learning.
Pengembangan sistem pengenalan maqam tilawah mujawwad secara keseluruhan dibagi menjadi dua tahap utama, yaitu pengembangan model pengenalan maqam dan pengembangan aplikasi mobile berbasis Android sebagai antarmuka sistem. Pada tahap pengembangan model pengenalan maqam tilawah mujawwad, proses-proses yang terlibat meliputi pengumpulan data, preprocessing data, ekstraksi fitur audio, pembuatan model, pelatihan model, dan evaluasi. Dalam penelitian ini, dilakukan berbagai skenario uji coba untuk mengevaluasi dan membandingkan performa berbagai konfigurasi model pengenalan maqam, yang mencakup preprocessing, kombinasi fitur, arsitektur model, dan hyperparameter yang digunakan.
Penelitian ini menghasilkan dataset tilawah mujawwad dengan total jumlah berkas audio sebanyak 269 dan total durasi sekitar 7,994 jam. Dataset ini memiliki variasi audio yang beragam, meliputi jenis kelamin, kualitas audio, dan tingkat keahlian dalam bacaan. Tahap preprocessing dilakukan dengan menghilangkan bagian yang silent dan membagi audio menjadi segmen-segmen dengan durasi yang sama menggunakan perpindahan frame tertentu (hop length). Ekstraksi fitur audio melibatkan penggunaan kombinasi fitur, seperti MFCC, ZCR, Chroma Feature, RMS Energy, Spectral Centroid, Spectral Bandwidth, dan Spectral Roll-Off. Model yang digunakan dalam penelitian ini adalah CNN, LSTM, CNN-LSTM, dan Deep ANN dengan input berupa nilai rata-rata. Hyperparameter tuning dilakukan untuk menemukan learning rate, optimizer, dan batch size yang optimal. Hasil evaluasi terhadap model menunjukkan bahwa performa terbaik didapatkan pada penggunaan hop length 5 dan semua kombinasi fitur digunakan, arsitektur Deep ANN dengan penggunaan fitur yang telah diagregasi menjadi nilai rata-rata untuk setiap segmen (1D), optimizer RMSprop, learning rate 0,0001, dan batch size 64, menghasilkan performa akurasi sebesar 0,969 dan F1-score sebesar 0,969. Penelitian ini juga menghasilkan aplikasi berbasis Android bernama Harmoni Qur’an, memungkinkan pengguna untuk mengakses dan memanfaatkan teknologi ini dengan mudah.
======================================================================
This research aims to develop a model for classifying maqams of Qur’anic recitations in the mujawwad style as well as an Android-based application as a model interface inference. A classification of maqams in mujawwad recitations is important for those who want to deepen the art of this recitation, but often face obstacles in understanding the different characteristics of the melody, rhythm and tone changes that are unique to each type of maqams. Apart from that, studying recitations requires special guidance from an Al-Qur'an teacher. Without the presence of a teacher, students will face difficulties in applying maqam which may not be appropriate. So, another approach is needed to overcome this problem, namely through the use of Deep Learning.
The overall development of the maqam recitation mujawwad recognition system is divided into two main stages, namely the development of the maqam classification model and the development of an Android-based mobile application as the system interface. At the development stage of the model for classifying maqam recitations in mujawwad style, the processes involved include data collection, data preprocessing, audio feature extraction, model building, model training, and evaluation. In this research, various test scenarios were carried out to evaluate and compare the performance of various maqam classification model configurations, which include preprocessing, feature combinations, model architecture, and hyperparameters used.
This research produced a dataset of mujawwad recitations with a total number of audio files of 269 and a total duration of around 7,994 hours. This dataset has a variety of audio variations, including gender, audio quality, and level of expertise in reading. The preprocessing stage is carried out by removing silent parts and dividing the audio into segments of the same duration using a certain hop length. Audio feature extraction involves using a combination of features, such as MFCC, ZCR, Chroma Feature, RMS Energy, Spectral Centroid, Spectral Bandwidth, and Spectral Roll-Off. The models used in this research are CNN, LSTM, CNN-LSTM, and Deep ANN with input in the form of average values. Hyperparameter tuning is carried out to find the optimal learning rate, optimizer and batch size values. The evaluation results of the model show that the best performance is obtained when using a hop length of 5 and all combinations of features are used, Deep ANN architecture uses features that have been aggregated into an average value for each segment (1D), RMSprop optimizer, learning rate 0.0001, batch size 64, and dropout 0.1, produces accuracy performance of 0.969 and F1-Score of 0.969. This research also produced an Android-based application called Harmoni Qur'an, allowing users to access and utilize this technology.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Android, Deep Learning, Ekstraksi Fitur Audio, Hyperparameter Tuning, Pengenalan Maqam Tilawah Mujawwad, Android, Audio Feature Extraction, Classification of Maqams in Mujawwad Qur’anic Recitation, Deep Learning, Hyperparameter Tuning.
Subjects:	Q Science > QA Mathematics > QA336 Artificial Intelligence Q Science > QA Mathematics > QA76.774.A53 Android Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) T Technology > T Technology (General) > T57.5 Data Processing
Divisions:	Faculty of Intelligent Electrical and Informatics Technology (ELECTICS) > Informatics Engineering > 55201-(S1) Undergraduate Thesis
Depositing User:	Arief Badrus Sholeh
Date Deposited:	02 Aug 2024 05:52
Last Modified:	02 Aug 2024 05:52
URI:	http://repository.its.ac.id/id/eprint/110258

Actions (login required)

View Item