KLASIFIKASI JENIS MAQAM PEMBACAAN AL-QURAN BERGAYA MUJAWWAD MENGGUNAKAN EKSTRAKSI FITUR MEL FREQUENCY CEPSTRAL COEFFICIENT (MFCC), DELTA, STATISTIK DAN NAÏVE BAYES CLASSIFIER

SHOLIHAH, NURIYATUS (2025) KLASIFIKASI JENIS MAQAM PEMBACAAN AL-QURAN BERGAYA MUJAWWAD MENGGUNAKAN EKSTRAKSI FITUR MEL FREQUENCY CEPSTRAL COEFFICIENT (MFCC), DELTA, STATISTIK DAN NAÏVE BAYES CLASSIFIER. Skripsi thesis, UPN Veteran Yogyakarta.

	Text 1_Skripsi Full_123210129_Nuriyatus Sholihah.pdf Restricted to Repository staff only Download (6MB)
	Text 2_Cover_123210129_Nuriyatus Sholihah.pdf Download (225kB)
	Text 3_Abstrak_123210129_Nuriyatus Sholihah.pdf Download (255kB)
	Text 4_Halaman Pengesahan Pembimbing dan Penguji_123210129_Nuriyatus Sholihah.pdf Download (1MB)
	Text 5_Daftar Isi_123210129_Nuriyatus Sholihah.pdf Download (258kB)
	Text 6_Daftar Pustaka_123210129_Nuriyatus Sholihah.pdf Download (231kB)

Abstract

ABSTRAK
Tilawah Al-Quran bergaya mujawwad merupakan salah satu bentuk seni yang
dianjurkan dalam Islam dan memiliki karakteristik khas berupa penggunaan maqam yang
kompleks. Setiap maqam memiliki ciri tersendiri dalam hal pola nada, tempo bacaan, dan
transisi antar tinggi nada atau cengkok yang khas, sehingga diperlukan pelatihan intensif
untuk dapat membedakannya dengan baik. Selama ini, proses pembelajaran maqam
dilakukan secara konvensional melalui talaqqi secara langsung bersama guru, yang tidak
mudah diakses oleh para pembelajar. Bagi pemula, membedakan antar maqam sering kali
menjadi tantangan tersendiri, bahkan pada tingkat lanjut pun masih ditemukan kesulitan
dalam mengenali nama maqam yang sedang dilantunkan. Kondisi ini menunjukkan
pentingnya dukungan teknologi sebagai alat bantu klasifikasi maqam secara otomatis untuk
mendukung proses evaluasi mandiri.
Penelitian ini menggunakan metode ekstraksi fitur Mel-Frequency Cepstral
Coefficient (MFCC) yang dikombinasikan dengan fitur turunan seperti delta dan delta-delta,
serta fitur statistik untuk menangkap karakteristik akustik dari tilawah mujawwad. Dataset
dikumpulkan secara mandiri dari berbagai channel YouTube tilawah, kemudian melalui
proses konversi, trimming, segmentasi, dan augmentasi hingga diperoleh 553 file audio
berdurasi sekitar 25 hingga 30 detik. Model klasifikasi dibangun menggunakan algoritma
Naïve Bayes Classifier, yang dikenal sederhana namun efektif untuk skala data kecil dengan
representasi fitur yang baik. Evaluasi dilakukan melalui beberapa skenario, termasuk
pengujian kombinasi fitur, variasi jumlah kelas maqam, dan pengujian kemampuan
generalisasi model terhadap data baru.
Hasil penelitian menunjukkan bahwa kombinasi fitur MFCC, delta, delta-delta, serta
statistik (mean, median, standard deviation, skewness, dan kurtosis) menghasilkan performa
terbaik dalam klasifikasi tujuh maqam dengan nilai macro average precision (MAP) sebesar
85,63%. Di antara fitur statistik tersebut, skewness menjadi fitur yang paling berkontribusi
dalam meningkatkan performa model. Bahkan, penggunaan subset fitur tertentu dapat
menghasilkan MAP tertinggi sebesar 90,27%, yang menegaskan pentingnya pemilihan fitur
yang tepat. Ketika jumlah kelas disederhanakan menjadi empat maqam, performa model
meningkat hingga 91,73%. Evaluasi terhadap 35 data baru juga menunjukkan bahwa model
berhasil mengenali enam maqam secara konsisten dengan total 30 data berhasil
diklasifikasikan dengan benar, sementara seluruh data maqam Jiharkah gagal
diklasifikasikan, yang menunjukkan keterbatasan kemampuan generalisasi model akibat
ketidakseimbangan distribusi data latih.
Kata Kunci: Klasifikasi Maqam, Tilawah Mujawwad, MFCC, Naïve Bayes Classifier,
Klasifikasi Audio
vii
ABSTRACT
Mujawwad-style Quranic recitation is considered one of the artistic forms promoted
in Islam and is distinguished by its intricate use of maqamat. Each maqam has unique
features, such as melodic patterns, reading tempo, and transitions between pitch levels
known as cengkok that require intensive training to recognize accurately. Traditionally,
maqam learning is carried out through talaqqi, a direct method of oral transmission from
teacher to student, which is often not easily accessible to many learners. For beginners,
distinguishing between maqamat can be especially challenging, and even at an advanced
level, identifying the specific maqam being recited can remain a difficult task. This situation
highlights the importance of technological support as a tool for automatic maqam
classification, thereby facilitating the self-evaluation process.
This study employs the Mel-Frequency Cepstral Coefficient (MFCC) feature
extraction method, combined with derivative features such as delta and delta-delta, and
statistical features, to capture the acoustic characteristics of mujawwad recitations. The
dataset was collected independently from various YouTube channels. Through a process of
conversion, trimming, segmentation, and augmentation, 553 audio files were obtained, each
with a duration of approximately 25 to 30 seconds. The classification model is built using
the Naïve Bayes Classifier algorithm, which is known to be a simple yet effective algorithm
for small data scales with good feature representation. The evaluation was conducted
through several scenarios, including testing feature combinations, variations in the number
of maqam classes, and assessing the model's generalization ability against new data.
The research findings indicate that a combination of MFCC features, delta, delta
delta, and statistical measures (mean, median, standard deviation, skewness, and kurtosis)
yields the best performance in classifying seven maqamat, with a macro average precision
(MAP) score of 85.63%. Among the statistical features, skewness contributed the most to
improving the model’s performance. Using a certain subset of features resulted in the highest
MAP of 90.27%, underscoring the importance of proper feature selection. When the number
of classes was simplified to four maqamat, the model’s performance increased to 91.73%.
Evaluation on 35 new data samples also showed that the model successfully identified six
maqamat consistently, with a total of 30 samples correctly classified. However, all samples
of the Jiharkah maqam failed to be classified, indicating limitations in the model’s
generalization ability due to an imbalance in the training data distribution.
Keywords: Maqam Classification, Mujawwad Recitation, MFCC, Naïve Bayes Classifier,
Audio Classification
viii

Item Type:	Tugas Akhir (Skripsi)
Uncontrolled Keywords:	Klasifikasi Maqam, Tilawah Mujawwad, MFCC, Naïve Bayes Classifier, Klasifikasi Audio
Subjek:	H Social Sciences > H Social Sciences (General)
Divisions:	Fakultas Teknik Industri > (S1) Informatika
Depositing User:	Eko Yuli
Date Deposited:	28 Oct 2025 07:10
Last Modified:	28 Oct 2025 07:10
URI:	http://eprints.upnyk.ac.id/id/eprint/45107

Actions (login required)

View Item