Date of Publication
2022
Document Type
Master's Thesis
Degree Name
Master of Science in Computer Science
Subject Categories
Computer Sciences
College
College of Computer Studies
Department/Unit
Computer Science
Thesis Advisor
Rafael A. Cabredo
Defense Panel Chair
Macario Cordel II
Defense Panel Member
Arturo Caronoñgan III
Rafael A. Cabredo
Abstract/Summary
Drum Track Generation is the problem of composing the rhythmic component of music. Drum Track Generation techniques are concerned with the composition of drum patterns given different parameters and input. Given the correct data and parameters, these techniques are capable of producing different types of drum patterns with varying styles and genre. Existing studies have proven the effectiveness of neural networks and Long Short-Term Memory (LSTM) models in generating varying drum outputs for various different purposes. However, most existing systems tend to generate outputs that lack structure and often become lost without a means of organization. Given that challenge, there exists an opportunity to generate drum pattern outputs using LSTM networks that exhibit a form of structure and organization. Presented in this study is a novel approach for generating structured drum tracks using LSTMs. In this study, a Markov Chain Model, LSTMs, and a novel architecture is used to generate the intended output. The technique presented in this study utilizes multiple databases of drum patterns sorted into the song structure segment they are classified as and a database of different song structure sequences. Evaluation studies were conducted and results indicate that the novel approach is able to generate drum tracks that are pleasant-sounding, perform better than previous work in terms of structure, and are able to be used in actual musical compositions.
Abstract Format
html
Language
English
Format
Electronic
Physical Description
372 leaves
Keywords
Composition (Music); Neural networks (Computer science)
Recommended Citation
Baarde, M. R. (2022). Structured drum track generation using long short-term memory networks. Retrieved from https://animorepository.dlsu.edu.ph/etdm_comsci/15
Upload Full Text
wf_yes
Embargo Period
2-26-2022