Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria

Daichi Sakaue, Takuma Otsuka, Katsutoshi Itoyama, Hiroshi G. Okuno

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

Abstract

We present a new Bayesian multipitch analyzer that dispenses with a precise optimization of parameter initialization or hyperparameters. Our method uses a new family of prior distribution, characteristic prior; it efficiently restricts the existence region of the latent variables, that is, the product of a conjugate prior and a characteristic function. The update formulas become a simple form that is actually suitable for Gibbs sampling. We construct characteristic priors of harmonic structures based on psychoacoustical and musical knowledge and apply them to nonnegative harmonic factorization. Experimental results improve 5.2 points in F-measure under a tough condition, random initialization with no hyperparameter optimization.

Original languageEnglish
Title of host publicationICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings
Pages226-230
Number of pages5
DOIs
Publication statusPublished - 2013 Oct 18
Externally publishedYes
Event2013 38th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2013 - Vancouver, BC
Duration: 2013 May 262013 May 31

Other

Other2013 38th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2013
CityVancouver, BC
Period13/5/2613/5/31

Fingerprint

Factorization
Sampling

Keywords

  • Bayesian analysis
  • harmonic clustering
  • multipitch estimation
  • nonnegative matrix factorization
  • overtone corpus

ASJC Scopus subject areas

  • Signal Processing
  • Software
  • Electrical and Electronic Engineering

Cite this

Sakaue, D., Otsuka, T., Itoyama, K., & Okuno, H. G. (2013). Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings (pp. 226-230). [6637642] https://doi.org/10.1109/ICASSP.2013.6637642

Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria. / Sakaue, Daichi; Otsuka, Takuma; Itoyama, Katsutoshi; Okuno, Hiroshi G.

ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2013. p. 226-230 6637642.

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

Sakaue, D, Otsuka, T, Itoyama, K & Okuno, HG 2013, Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria. in ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings., 6637642, pp. 226-230, 2013 38th IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2013, Vancouver, BC, 13/5/26. https://doi.org/10.1109/ICASSP.2013.6637642
Sakaue D, Otsuka T, Itoyama K, Okuno HG. Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria. In ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2013. p. 226-230. 6637642 https://doi.org/10.1109/ICASSP.2013.6637642
Sakaue, Daichi ; Otsuka, Takuma ; Itoyama, Katsutoshi ; Okuno, Hiroshi G. / Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria. ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings. 2013. pp. 226-230
@inproceedings{70304ad805424c8f98809570f6e8d5d0,
title = "Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria",
abstract = "We present a new Bayesian multipitch analyzer that dispenses with a precise optimization of parameter initialization or hyperparameters. Our method uses a new family of prior distribution, characteristic prior; it efficiently restricts the existence region of the latent variables, that is, the product of a conjugate prior and a characteristic function. The update formulas become a simple form that is actually suitable for Gibbs sampling. We construct characteristic priors of harmonic structures based on psychoacoustical and musical knowledge and apply them to nonnegative harmonic factorization. Experimental results improve 5.2 points in F-measure under a tough condition, random initialization with no hyperparameter optimization.",
keywords = "Bayesian analysis, harmonic clustering, multipitch estimation, nonnegative matrix factorization, overtone corpus",
author = "Daichi Sakaue and Takuma Otsuka and Katsutoshi Itoyama and Okuno, {Hiroshi G.}",
year = "2013",
month = "10",
day = "18",
doi = "10.1109/ICASSP.2013.6637642",
language = "English",
isbn = "9781479903566",
pages = "226--230",
booktitle = "ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings",

}

TY - GEN

T1 - Initialization-robust Bayesian multipitch analyzer based on psychoacoustical and musical criteria

AU - Sakaue, Daichi

AU - Otsuka, Takuma

AU - Itoyama, Katsutoshi

AU - Okuno, Hiroshi G.

PY - 2013/10/18

Y1 - 2013/10/18

N2 - We present a new Bayesian multipitch analyzer that dispenses with a precise optimization of parameter initialization or hyperparameters. Our method uses a new family of prior distribution, characteristic prior; it efficiently restricts the existence region of the latent variables, that is, the product of a conjugate prior and a characteristic function. The update formulas become a simple form that is actually suitable for Gibbs sampling. We construct characteristic priors of harmonic structures based on psychoacoustical and musical knowledge and apply them to nonnegative harmonic factorization. Experimental results improve 5.2 points in F-measure under a tough condition, random initialization with no hyperparameter optimization.

AB - We present a new Bayesian multipitch analyzer that dispenses with a precise optimization of parameter initialization or hyperparameters. Our method uses a new family of prior distribution, characteristic prior; it efficiently restricts the existence region of the latent variables, that is, the product of a conjugate prior and a characteristic function. The update formulas become a simple form that is actually suitable for Gibbs sampling. We construct characteristic priors of harmonic structures based on psychoacoustical and musical knowledge and apply them to nonnegative harmonic factorization. Experimental results improve 5.2 points in F-measure under a tough condition, random initialization with no hyperparameter optimization.

KW - Bayesian analysis

KW - harmonic clustering

KW - multipitch estimation

KW - nonnegative matrix factorization

KW - overtone corpus

UR - http://www.scopus.com/inward/record.url?scp=84890443689&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84890443689&partnerID=8YFLogxK

U2 - 10.1109/ICASSP.2013.6637642

DO - 10.1109/ICASSP.2013.6637642

M3 - Conference contribution

AN - SCOPUS:84890443689

SN - 9781479903566

SP - 226

EP - 230

BT - ICASSP, IEEE International Conference on Acoustics, Speech and Signal Processing - Proceedings

ER -