PENGEMBANGAN MODEL CONVLSTM UNTUK KLASIFIKASI AUDIO GAMELAN BALI
Abstract
Abstrak. Musik tradisional seperti Gamelan Bali menghadapi tantangan pelestarian di era digital. Proses identifikasi manual yang lambat dan subjektif menghambat upaya pengarsipan skala besar. Penelitian ini mengusulkan solusi melalui pengembangan model klasifikasi audio otomatis menggunakan arsitektur Convolutional Long Short-Term Memory. Tujuan penelitian adalah untuk merancang, melatih, dan mengevaluasi model yang mampu mengklasifikasikan lima Gamelan Bali: Angklung, Baleganjur, Gong Gede, Gong Kebyar, dan Semar Pegulingan. Metode penelitian ini mencakup beberapa tahapan, mulai dari pengumpulan dan validasi dataset oleh ahli, pra-pemrosesan audio dengan segmentasi durasi tetap dan denoising, hingga ekstraksi fitur gabungan (MFCC, Chroma, Spectral Contrast). Beberapa skenario pelatihan dieksperimenkan untuk menemukan konfigurasi optimal. Hasil penelitian menunjukkan bahwa model terbaik, yang dilatih menggunakan audio 10 detik dengan teknik transformasi data, berhasil mencapai akurasi validasi sebesar 75%. Analisis lebih lanjut mengungkap bahwa model menunjukkan performa baik pada kelas dengan karakteristik unik seperti Baleganjur (F1-Score 0.91), namun menghadapi tantangan pada kelas dengan tumpang tindih akustik seperti Gong Kebyar (F1-Score 0.50). Disimpulkan bahwa arsitektur ConvLSTM cukup dalam klasifikasi audio Gamelan Bali namun tetap perlu ditingkatkan.
Abstract. Traditional music such as Balinese Gamelan faces preservation challenges in the digital age. Slow and subjective manual identification processes hinder large-scale archiving efforts. This study proposes a solution through the development of an automatic audio classification model using a Convolutional Long Short-Term Memory architecture. The objective of this study is to design, train, and evaluate a model capable of classifying five types of Balinese Gamelan: Angklung, Baleganjur, Gong Gede, Gong Kebyar, and Semar Pegulingan. The research method includes several stages, starting from data collection and validation by experts, audio pre-processing with fixed duration segmentation and denoising, to combined feature extraction (MFCC, Chroma, Spectral Contrast). Several training scenarios were experimented with to find the optimal configuration. The results show that the best model, trained using 10-second audio clips with data transformation techniques, achieved a validation accuracy of 75%. Further analysis revealed that the model performed well on classes with unique characteristics such as Baleganjur (F1-Score 0.91), but faced challenges on classes with acoustic overlap such as Gong Kebyar (F1-Score 0.50). It was concluded that the ConvLSTM architecture is sufficient for Balinese Gamelan audio classification but still needs improvement.
Downloads
References
H. Guo, J. Wang, X. Xu, C. C. Hung, dan S. An, “A Hybrid Deep Learning Model for Automatic Classification of Cantonese Opera,” Sustainability, vol. 14, no. 5, hlm. 2923, Mar. 2022.
A. M. Tarsitani dan G. G. Acosta, “Artificial intelligence in music,” Revista de la Asociación Argentina de Musicología, vol. 27, no. 1, hlm. 1-22, 2021.
K. L. Gee, “The Gamelan,” BYU Physics and Astronomy, 2022. [Online]. Tersedia: https://physics.byu.edu/faculty/gee/gamelan.
W. Vitale, “Balinese Gamelan Semar Pagulingan,” Gamelan.org, 2002. [Online]. Tersedia: https://www.gamelan.org/ensembles/saih-pitu.
L. A. Kurniawati, A. T. Hidayatno, dan I. P. H. S. Putra, “Deep Learning for Multi-Structured Javanese Gamelan Note Generator,” Jurnal Ilmiah Teknologi dan Informasi Asia, vol. 17, no. 2, hlm. 119-128, Agu. 2023.
Z. Shi, H. Wang, dan Y. Li, “An Attention-based Conv-LSTM for Spatiotemporal Flow Prediction,” arXiv preprint arXiv:2006.08520, 2020.
A. Kaipainen, “Challenges of Music in the Digital Era,” THEORIA: An International Journal for Theory, History, and Foundations of Science, vol. 34, no. 1, hlm. 121-125, Jan. 2019.
F. M. Dewadi, "Pembelajaran dan pengenalan musik dalam melatih daya ingat anak," JECED: Journal of Early Childhood Education and Development, vol. 3, no. 1, hlm. 15-23, 2021.
J. S. Atmadja, "Membangun game menggunakan metode agile development," Medium, 30 Apr. 2020. [Online]. Tersedia: https://medium.com/javanlabs/membangun-game-menggunakan-metode-agile-development-a661cb48fe81.
A. Haddid, M. Q. Aufar, A. Ramadhan, dan Y. Pratama, "Perancangan game edukasi menebak lirik lagu-lagu daerah berbasis Unity 2D untuk platform Android," TIN: Terapan Informatika Nusantara, vol. 4, no. 2, hlm. 140-151, 2023.
I. N. R. Artana, "Balinese Gamelan Learning Process During The Covid-19 Pandemic," LULIK: Jurnal Ilmiah Kreatifitas, Seni, dan Budaya, vol. 11, no. 2, hlm. 106-115, 2022.
I. G. G. Surya, I. N. D. Putra, dan N. M. R. A. Gelgel, "The Balinese gamelan phenomenon in digital simulation," E-Journal of Cultural Studies, vol. 17, no. 3, hlm. 1-10, 2024.
G. A. Pratama, E. Y. Puspaningrum, dan H. Maulana, "Convolutional Neural Network dan Faster Region Convolutional Neural Network Untuk Klasifikasi Kualitas Biji Kopi Arabika," Jurnal Informatika dan Teknik Elektro Terapan (JITET), vol. 12, no. 3, 2024.
A. Shvets, I. Vankov, dan D. Serokutov, “Spiking ConvLSTM for semantic music generation,” arXiv preprint arXiv:1911.11447, 2019.
A. Telmem, “Speech recognition based on convolutional neural networks and MFCC algorithm,” International Journal of Advanced Computer Science and Applications, vol. 12, no. 3, 2021.
G. K. Birajdar dan M. Patil, “Speech/music classification using visual and spectral chromagram features,” Journal of Ambient Intelligence and Humanized Computing, vol. 11, no. 1, pp. 329–347, 2020.
D. N. Jiang, L. Lu, H. J. Zhang, J. H. Tao, dan L. H. Cai, “Music type classification by spectral contrast feature,” dalam Proc. IEEE Int. Conf. Multimedia and Expo, 2002, vol. 1, pp. 113–116.
A. Fahrezi, F. N. Salam, G. M. Ibrahim, R. Rahman, dan A. Saifudin, "Pengujian black box testing pada aplikasi inventori barang berbasis web di PT. AINO Indonesia," Jurnal Teknologi dan Sistem Informasi (JTSI), vol. 1, no. 1, hlm. 61-66, 2022.
Ž. Vujović, “Classification model evaluation metrics,” International Journal of Advanced Computer Science and Applications, vol. 12, no. 6, 2021.
S. Venkatesh, D. Moffat, dan E. R. Miranda, “You only hear once: A YOLO like algorithm for audio segmentation and sound event detection,” Applied Sciences, vol. 12, no. 7, p. 3293, 2022.
J. Salamon dan J. P. Bello, “Deep convolutional neural networks and data augmentation for environmental sound classification,” IEEE Signal Processing Letters, vol. 24, no. 3, pp. 279-283, 2017.
K. West dan S. Cox, “Finding an optimal segmentation for audio genre classification,” dalam Proc. 25th Int. Society for Music Information Retrieval Conf., ISMIR 2024, 2024.
J. Zheng, M. Cao, dan C. Zhang, “ICKAN: A deep musical instrument classification model incorporating Kolmogorov-Arnold network,” Scientific Reports, vol. 14, no. 1, p. 9493, 2024.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
