HMM analysis of musical structure

Identification of latent variables through topology-sensitive model selection

Panayotis Mavromatis

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

Abstract

Hidden Markov Models (HMMs) have been successfully employed in the exploration and modeling of musical structure, with applications in Music Information Retrieval. This paper focuses on an aspect of HMM training that remains relatively unexplored in musical applications, namely the determination of HMM topology. We demonstrate that this complex problem can be effectively addressed through search over model topology space, conducted by HMM state merging and/or splitting. Once successfully identified, the HMM topology that is optimal with respect to a given data set can help identify hidden (latent) variables that are important in shaping the data set's visible structure. These variables are identified by suitable interpretation of the HMM states for the selected topology. As an illustration, we present two case studies that successfully tackle two classic problems in music computation, namely (i) algorithmic statistical segmentation and (ii) meter induction from a sequence of durational patterns.

Original languageEnglish (US)
Title of host publicationMathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings
Pages205-217
Number of pages13
Volume38
DOIs
StatePublished - 2009

Publication series

NameCommunications in Computer and Information Science
Volume38
ISSN (Print)18650929

Fingerprint

Hidden Markov models
Identification (control systems)
Topology
Information retrieval
Merging

ASJC Scopus subject areas

  • Computer Science(all)

Cite this

Mavromatis, P. (2009). HMM analysis of musical structure: Identification of latent variables through topology-sensitive model selection. In Mathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings (Vol. 38, pp. 205-217). (Communications in Computer and Information Science; Vol. 38). https://doi.org/10.1007/978-3-642-02394-1_19

HMM analysis of musical structure : Identification of latent variables through topology-sensitive model selection. / Mavromatis, Panayotis.

Mathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings. Vol. 38 2009. p. 205-217 (Communications in Computer and Information Science; Vol. 38).

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

Mavromatis, P 2009, HMM analysis of musical structure: Identification of latent variables through topology-sensitive model selection. in Mathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings. vol. 38, Communications in Computer and Information Science, vol. 38, pp. 205-217. https://doi.org/10.1007/978-3-642-02394-1_19
Mavromatis P. HMM analysis of musical structure: Identification of latent variables through topology-sensitive model selection. In Mathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings. Vol. 38. 2009. p. 205-217. (Communications in Computer and Information Science). https://doi.org/10.1007/978-3-642-02394-1_19
Mavromatis, Panayotis. / HMM analysis of musical structure : Identification of latent variables through topology-sensitive model selection. Mathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings. Vol. 38 2009. pp. 205-217 (Communications in Computer and Information Science).
@inproceedings{76709b5e41ff4b77998fd0562be3ed5b,
title = "HMM analysis of musical structure: Identification of latent variables through topology-sensitive model selection",
abstract = "Hidden Markov Models (HMMs) have been successfully employed in the exploration and modeling of musical structure, with applications in Music Information Retrieval. This paper focuses on an aspect of HMM training that remains relatively unexplored in musical applications, namely the determination of HMM topology. We demonstrate that this complex problem can be effectively addressed through search over model topology space, conducted by HMM state merging and/or splitting. Once successfully identified, the HMM topology that is optimal with respect to a given data set can help identify hidden (latent) variables that are important in shaping the data set's visible structure. These variables are identified by suitable interpretation of the HMM states for the selected topology. As an illustration, we present two case studies that successfully tackle two classic problems in music computation, namely (i) algorithmic statistical segmentation and (ii) meter induction from a sequence of durational patterns.",
author = "Panayotis Mavromatis",
year = "2009",
doi = "10.1007/978-3-642-02394-1_19",
language = "English (US)",
isbn = "9783642023934",
volume = "38",
series = "Communications in Computer and Information Science",
pages = "205--217",
booktitle = "Mathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings",

}

TY - GEN

T1 - HMM analysis of musical structure

T2 - Identification of latent variables through topology-sensitive model selection

AU - Mavromatis, Panayotis

PY - 2009

Y1 - 2009

N2 - Hidden Markov Models (HMMs) have been successfully employed in the exploration and modeling of musical structure, with applications in Music Information Retrieval. This paper focuses on an aspect of HMM training that remains relatively unexplored in musical applications, namely the determination of HMM topology. We demonstrate that this complex problem can be effectively addressed through search over model topology space, conducted by HMM state merging and/or splitting. Once successfully identified, the HMM topology that is optimal with respect to a given data set can help identify hidden (latent) variables that are important in shaping the data set's visible structure. These variables are identified by suitable interpretation of the HMM states for the selected topology. As an illustration, we present two case studies that successfully tackle two classic problems in music computation, namely (i) algorithmic statistical segmentation and (ii) meter induction from a sequence of durational patterns.

AB - Hidden Markov Models (HMMs) have been successfully employed in the exploration and modeling of musical structure, with applications in Music Information Retrieval. This paper focuses on an aspect of HMM training that remains relatively unexplored in musical applications, namely the determination of HMM topology. We demonstrate that this complex problem can be effectively addressed through search over model topology space, conducted by HMM state merging and/or splitting. Once successfully identified, the HMM topology that is optimal with respect to a given data set can help identify hidden (latent) variables that are important in shaping the data set's visible structure. These variables are identified by suitable interpretation of the HMM states for the selected topology. As an illustration, we present two case studies that successfully tackle two classic problems in music computation, namely (i) algorithmic statistical segmentation and (ii) meter induction from a sequence of durational patterns.

UR - http://www.scopus.com/inward/record.url?scp=67649989633&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=67649989633&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-02394-1_19

DO - 10.1007/978-3-642-02394-1_19

M3 - Conference contribution

SN - 9783642023934

VL - 38

T3 - Communications in Computer and Information Science

SP - 205

EP - 217

BT - Mathematics and Computation in Music: Second International Conference, MCM 2009, John Clough Memorial Conference, Proceedings

ER -