Implementasi Metode Long Short-Term Memory Dalam Pembuatan Musik Sintetis

Muzli, Alfarabi (2023) Implementasi Metode Long Short-Term Memory Dalam Pembuatan Musik Sintetis. Other thesis, Institut Teknologi Sepuluh Nopember.

Text 05111940000039-Undergraduate_Thesis.pdf - Accepted Version Restricted to Repository staff only until 1 September 2025. Download (4MB) | Request a copy

Abstract

Penggunaan teknologi dalam bidang seni telah membuka berbagai peluang baru untuk meningkatkan kinerja dan kreativitas. Dalam industri musik, perkembangan pesat kecerdasan buatan telah memberikan kemungkinan baru dalam pembuatan musik. Namun, keberadaan teknologi ini juga memiliki implikasi terhadap penggunaan musik dan proses transkripsi manual. Musik menjadi sumber daya penting bagi para musisi, telah mengalami perubahan dari bentuk manuskrip tangan menjadi versi digital yang dapat diakses melalui internet. Pelatihan pembuatan musik menggunakan Recurrent Neural Network yaitu Long Short-Term Memory dilakukan dengan dataset terdiri dari tiga orang komponis ternama dalam bidang musik klasik. Dataset ini dibagi menjadi empat bagian. Selain itu, ada dua skenario yang akan diuji, dan hasilnya akan dibandingkan dengan model lain yaitu Gated Recurrent Unit. Untuk pengolahan data musik, digunakan library python bernama Music21. Model dievaluasi dengan empat cara yaitu dengan melihat loss pelatihan model, persebaran statistik notasi, perhitungan kesamaan dan penilaian dari ahli musik. Pada skenario kedua, model LSTM menunjukkan hasil yang lebih baik dibandingkan dengan skenario pertama, dan juga lebih unggul daripada model GRU.
==================================================================================================================================
The use of technology in the arts has opened up new opportunities to enhance performance and creativity. In the music industry, the rapid development of artificial intelligence has provided new possibilities in music creation. However, it also has implications for the use of music and the manual transcription process. Music, as an important resource for musicians, have undergone a change from a hand manuscript form to a digital version that can be accessed through the internet. Using Recurrent Neural Network, namely Long Short-Term Memory to perform music generation training with a dataset consisting of three well-known composers in the field of classical music which is divided into four datasets and two scenarios and will be compared with another model namely Gated Recurrent Unit using Music21 as a python library for music data processing. The model was evaluated in four ways: looking at the training loss of the model, the distribution of notation statistics, the similarity calculation, and the scoring of music experts.. The results show that in the second scenario of the LSTM model, it gets good results compared to the first scenario and better than the GRU model.

Item Type:	Thesis (Other)
Uncontrolled Keywords:	Teknologi Dalam Seni, Musik, Recurrent Neural Network, Long Short-Term Memory, Musik Klasik, Music21; Technology in Arts, Music, Recurrent Neural Network, Long Short-Term Memory, Classical Music, Music21
Subjects:	M Music and Books on Music > M Music Q Science > Q Science (General) > Q325.5 Machine learning. Support vector machines. Q Science > QA Mathematics > QA76.87 Neural networks (Computer Science) R Medicine > R Medicine (General) > R858 Deep Learning 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:	Alfarabi Muzli
Date Deposited:	15 Sep 2023 07:37
Last Modified:	15 Sep 2023 07:37
URI:	http://repository.its.ac.id/id/eprint/102063

Actions (login required)

View Item