Machine learning for music genre: multifaceted review and experimentation with audioset
- PDF / 1,262,605 Bytes
- 31 Pages / 439.642 x 666.49 pts Page_size
- 26 Downloads / 166 Views
Machine learning for music genre: multifaceted review and experimentation with audioset Jaime Ram´ırez1
· M. Julia Flores1
Received: 23 July 2019 / Revised: 4 October 2019 / Accepted: 9 October 2019 / © Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract Music genre classification is one of the sub-disciplines of music information retrieval (MIR) with growing popularity among researchers, mainly due to the already open challenges. Although research has been prolific in terms of number of published works, the topic still suffers from a problem in its foundations: there is no clear and formal definition of what genre is. Music categorizations are vague and unclear, suffering from human subjectivity and lack of agreement. In its first part, this paper offers a survey trying to cover the many different aspects of the matter. Its main goal is give the reader an overview of the history and the current state-of-the-art, exploring techniques and datasets used to the date, as well as identifying current challenges, such as this ambiguity of genre definitions or the introduction of human-centric approaches. The paper pays special attention to new trends in machine learning applied to the music annotation problem. Finally, we also include a music genre classification experiment that compares different machine learning models using Audioset. Keywords Machine learning · Datasets · Music information retrieval · Classification algorithms · Music · Feed-forward neural networks
1 Introduction Music information retrieval (MIR) is an interdisciplinary field which covers different aspects concerning the extraction of information from music (Downie 2003), from sociological and musicological aspects to recommender systems, music generators or annotators (Liem et al. 2013; Kitahara 2017). Recent advances in machine learning (ML) models and artificial intelligence (AI) are replacing traditional approaches in MIR based sometimes on signal processing (L¨angkvist et al. 2014) and generating more accurate results. Jaime Ram´ırez
[email protected] M. Julia Flores [email protected] 1
Computing Systems Department, UCLM, Albacete, Spain
Journal of Intelligent Information Systems
One of the sub-problems of the music annotation domain exploring these advances is music genre classification (MGC). Historically, acoustic and sound characteristics have been the main features to consider when performing genre classification. For example, Jazz is usually characterized by swing rhythms, improvisation and instruments such as piano and trumpet, whereas Electronic music can be identified by the use of synthesizers. It is reasonable to think that, if provided with raw signals and acoustic characteristics, genre classification could be more precise. However, this approach has been only partially effective. Although genre is a crucial descriptor, not only for MGC, but for many problems related, its definition is unclear. The categorization of a piece of music into a musical genre is often subject to different human
Data Loading...
