Convolution-based classification of audio and symbolic representations of music

Gissel Velarde, Carlos Cancino Chacón, David Meredith, Tillman Weyde, Maarten Grachten

Research output: Contribution to journalJournal articleResearchpeer-review

Abstract

We present a novel convolution-based method for classification of audio and symbolic representations of music, which we apply to classification of music by style. Pieces of music are first sampled to pitch–time representations (piano-rolls or spectrograms) and then convolved with a Gaussian filter, before being classified by a support vector machine or by k-nearest neighbours in an ensemble of classifiers. On the well-studied task of discriminating between string quartet movements by Haydn and Mozart, we obtain accuracies that equal the state of the art on two data-sets. However, in multi-class composer identification, methods specialised for classifying symbolic representations of music are more effective. We also performed experiments on symbolic representations, synthetic audio and two different recordings of The Well-Tempered Clavier by J. S. Bach to study the method’s capacity to distinguish preludes from fugues. Our experimental results show that our approach performs similarly on symbolic representations, synthetic audio and audio recordings, setting our method apart from most previous studies that have been designed for use with either audio or symbolic data, but not both.
Original languageEnglish
JournalJournal of New Music Research
Volume47
Issue number3
Pages (from-to)191-205
Number of pages15
ISSN0929-8215
DOIs
Publication statusUnpublished - 2019

Fingerprint

Music
The Well-tempered Clavier
Nearest Neighbor
Joseph Haydn
Composer
Classifier
String Quartet
Filter
Support Vector Machine
Prelude
Audio Recordings
Johann Sebastian Bach
Experiment
Wolfgang Amadeus Mozart
Ensemble

Keywords

  • Music analysis
  • machine learning
  • convolution
  • composer recognition
  • genre recognition

Cite this

Velarde, Gissel ; Cancino Chacón, Carlos ; Meredith, David ; Weyde, Tillman ; Grachten, Maarten. / Convolution-based classification of audio and symbolic representations of music. In: Journal of New Music Research. 2019 ; Vol. 47, No. 3. pp. 191-205.
@article{91cdb7d1f3bb4e00831e59121e88b03b,
title = "Convolution-based classification of audio and symbolic representations of music",
abstract = "We present a novel convolution-based method for classification of audio and symbolic representations of music, which we apply to classification of music by style. Pieces of music are first sampled to pitch–time representations (piano-rolls or spectrograms) and then convolved with a Gaussian filter, before being classified by a support vector machine or by k-nearest neighbours in an ensemble of classifiers. On the well-studied task of discriminating between string quartet movements by Haydn and Mozart, we obtain accuracies that equal the state of the art on two data-sets. However, in multi-class composer identification, methods specialised for classifying symbolic representations of music are more effective. We also performed experiments on symbolic representations, synthetic audio and two different recordings of The Well-Tempered Clavier by J. S. Bach to study the method’s capacity to distinguish preludes from fugues. Our experimental results show that our approach performs similarly on symbolic representations, synthetic audio and audio recordings, setting our method apart from most previous studies that have been designed for use with either audio or symbolic data, but not both.",
keywords = "Music analysis, machine learning, convolution, composer recognition, genre recognition",
author = "Gissel Velarde and {Cancino Chac{\'o}n}, Carlos and David Meredith and Tillman Weyde and Maarten Grachten",
year = "2019",
doi = "10.1080/09298215.2018.1458885",
language = "English",
volume = "47",
pages = "191--205",
journal = "Journal of New Music Research",
issn = "0929-8215",
publisher = "Routledge",
number = "3",

}

Convolution-based classification of audio and symbolic representations of music. / Velarde, Gissel; Cancino Chacón, Carlos; Meredith, David; Weyde, Tillman; Grachten, Maarten.

In: Journal of New Music Research, Vol. 47, No. 3, 2019, p. 191-205.

Research output: Contribution to journalJournal articleResearchpeer-review

TY - JOUR

T1 - Convolution-based classification of audio and symbolic representations of music

AU - Velarde, Gissel

AU - Cancino Chacón, Carlos

AU - Meredith, David

AU - Weyde, Tillman

AU - Grachten, Maarten

PY - 2019

Y1 - 2019

N2 - We present a novel convolution-based method for classification of audio and symbolic representations of music, which we apply to classification of music by style. Pieces of music are first sampled to pitch–time representations (piano-rolls or spectrograms) and then convolved with a Gaussian filter, before being classified by a support vector machine or by k-nearest neighbours in an ensemble of classifiers. On the well-studied task of discriminating between string quartet movements by Haydn and Mozart, we obtain accuracies that equal the state of the art on two data-sets. However, in multi-class composer identification, methods specialised for classifying symbolic representations of music are more effective. We also performed experiments on symbolic representations, synthetic audio and two different recordings of The Well-Tempered Clavier by J. S. Bach to study the method’s capacity to distinguish preludes from fugues. Our experimental results show that our approach performs similarly on symbolic representations, synthetic audio and audio recordings, setting our method apart from most previous studies that have been designed for use with either audio or symbolic data, but not both.

AB - We present a novel convolution-based method for classification of audio and symbolic representations of music, which we apply to classification of music by style. Pieces of music are first sampled to pitch–time representations (piano-rolls or spectrograms) and then convolved with a Gaussian filter, before being classified by a support vector machine or by k-nearest neighbours in an ensemble of classifiers. On the well-studied task of discriminating between string quartet movements by Haydn and Mozart, we obtain accuracies that equal the state of the art on two data-sets. However, in multi-class composer identification, methods specialised for classifying symbolic representations of music are more effective. We also performed experiments on symbolic representations, synthetic audio and two different recordings of The Well-Tempered Clavier by J. S. Bach to study the method’s capacity to distinguish preludes from fugues. Our experimental results show that our approach performs similarly on symbolic representations, synthetic audio and audio recordings, setting our method apart from most previous studies that have been designed for use with either audio or symbolic data, but not both.

KW - Music analysis

KW - machine learning

KW - convolution

KW - composer recognition

KW - genre recognition

U2 - 10.1080/09298215.2018.1458885

DO - 10.1080/09298215.2018.1458885

M3 - Journal article

VL - 47

SP - 191

EP - 205

JO - Journal of New Music Research

JF - Journal of New Music Research

SN - 0929-8215

IS - 3

ER -