Meter Detection in Symbolic Music Using a Lexicalized PCFG

Andrew McLeod, Mark Steedman

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This work proposes a lexicalized probabilistic context free grammar designed for meter detection, an integral component of automatic music transcription. The grammar uses rhythmic cues to align a given musical piece with learned metrical stress patterns. Lexicalization breaks the standard PCFG assumption of independence of production, and thus, our grammar can model the more complex rhythmic dependencies which are present in musical compositions.
Using a metric we propose for the task, we show that our grammar outperforms baseline methods when run on symbolic music input which has been hand-aligned to a tatum. We also show that the grammar outperforms an existing
method when run with automatically-aligned symbolic music data as input. The code for our grammar is available at https://github.com/apmcleod/met-detection.
Original languageEnglish
Title of host publicationProceedings of the 14th Sound and Music Computing Conference
PublisherAalto University
Pages373-379
Number of pages7
ISBN (Print)978-952-60-3729-5
Publication statusPublished - 8 Jul 2017
Event14th Sound and Music Computing Conference - Espoo, Finland
Duration: 5 Jul 20178 Jul 2017
http://smc2017.aalto.fi/

Publication series

NameProceedings of the 14th Sound and Music Computing Conference 2017
PublisherAalto University
ISSN (Electronic)2518-3672

Conference

Conference14th Sound and Music Computing Conference
Abbreviated titleSMC 2017
CountryFinland
CityEspoo
Period5/07/178/07/17
Internet address

Fingerprint Dive into the research topics of 'Meter Detection in Symbolic Music Using a Lexicalized PCFG'. Together they form a unique fingerprint.

Cite this