Pitch Estimation and Tracking with Harmonic Emphasis On The Acoustic Spectrum

Publikation: Bidrag til tidsskriftKonferenceartikel i tidsskriftForskningpeer review

4 Citationer (Scopus)
436 Downloads (Pure)

Resumé

In this paper, we use unconstrained frequency estimates (UFEs) from a noisy harmonic signal and propose two methods to estimate and track the pitch over time. We assume that the UFEs are multivariate-normally-distributed random variables, and derive a maximum likelihood (ML) pitch estimator by maximizing the likelihood of the UFEs over short time-intervals. As the main contribution of this paper, we propose two state-space representations to model the pitch continuity, and, accordingly, we propose two Bayesian methods, namely a hidden Markov model and a Kalman filter. These methods are designed to optimally use the correlations in the consecutive pitch values, where the past pitch estimates are used to recursively update the prior distribution for the pitch variable. We perform experiments using synthetic data as well as a noisy speech recording, and show that the Bayesian methods provide more accurate estimates than the corresponding ML methods.
OriginalsprogEngelsk
TidsskriftI E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings
Sider (fra-til)4330-4334
ISSN1520-6149
DOI
StatusUdgivet - apr. 2015
Begivenhed40th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2015 - Brisbane, Australien
Varighed: 19 apr. 201524 apr. 2015
Konferencens nummer: 2015

Konference

Konference40th IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) 2015
Nummer2015
LandAustralien
ByBrisbane
Periode19/04/201524/04/2015

Fingerprint

Maximum likelihood
Acoustics
Hidden Markov models
Random variables
Kalman filters
Experiments

Citer dette

@inproceedings{a5ddf69b88fa4fcab0b4225c17dfe074,
title = "Pitch Estimation and Tracking with Harmonic Emphasis On The Acoustic Spectrum",
abstract = "In this paper, we use unconstrained frequency estimates (UFEs) from a noisy harmonic signal and propose two methods to estimate and track the pitch over time. We assume that the UFEs are multivariate-normally-distributed random variables, and derive a maximum likelihood (ML) pitch estimator by maximizing the likelihood of the UFEs over short time-intervals. As the main contribution of this paper, we propose two state-space representations to model the pitch continuity, and, accordingly, we propose two Bayesian methods, namely a hidden Markov model and a Kalman filter. These methods are designed to optimally use the correlations in the consecutive pitch values, where the past pitch estimates are used to recursively update the prior distribution for the pitch variable. We perform experiments using synthetic data as well as a noisy speech recording, and show that the Bayesian methods provide more accurate estimates than the corresponding ML methods.",
keywords = "Harmonic signal, frequency estimate, pitch estimation, Bayesian filter, Kalman filter",
author = "Sam Karimian-Azari and Nasser Mohammadiha and Jensen, {Jesper Rindom} and Christensen, {Mads Gr{\ae}sb{\o}ll}",
year = "2015",
month = "4",
doi = "10.1109/ICASSP.2015.7178788",
language = "English",
pages = "4330--4334",
journal = "I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings",
issn = "1520-6149",
publisher = "IEEE Signal Processing Society",

}

TY - GEN

T1 - Pitch Estimation and Tracking with Harmonic Emphasis On The Acoustic Spectrum

AU - Karimian-Azari, Sam

AU - Mohammadiha, Nasser

AU - Jensen, Jesper Rindom

AU - Christensen, Mads Græsbøll

PY - 2015/4

Y1 - 2015/4

N2 - In this paper, we use unconstrained frequency estimates (UFEs) from a noisy harmonic signal and propose two methods to estimate and track the pitch over time. We assume that the UFEs are multivariate-normally-distributed random variables, and derive a maximum likelihood (ML) pitch estimator by maximizing the likelihood of the UFEs over short time-intervals. As the main contribution of this paper, we propose two state-space representations to model the pitch continuity, and, accordingly, we propose two Bayesian methods, namely a hidden Markov model and a Kalman filter. These methods are designed to optimally use the correlations in the consecutive pitch values, where the past pitch estimates are used to recursively update the prior distribution for the pitch variable. We perform experiments using synthetic data as well as a noisy speech recording, and show that the Bayesian methods provide more accurate estimates than the corresponding ML methods.

AB - In this paper, we use unconstrained frequency estimates (UFEs) from a noisy harmonic signal and propose two methods to estimate and track the pitch over time. We assume that the UFEs are multivariate-normally-distributed random variables, and derive a maximum likelihood (ML) pitch estimator by maximizing the likelihood of the UFEs over short time-intervals. As the main contribution of this paper, we propose two state-space representations to model the pitch continuity, and, accordingly, we propose two Bayesian methods, namely a hidden Markov model and a Kalman filter. These methods are designed to optimally use the correlations in the consecutive pitch values, where the past pitch estimates are used to recursively update the prior distribution for the pitch variable. We perform experiments using synthetic data as well as a noisy speech recording, and show that the Bayesian methods provide more accurate estimates than the corresponding ML methods.

KW - Harmonic signal

KW - frequency estimate

KW - pitch estimation

KW - Bayesian filter

KW - Kalman filter

U2 - 10.1109/ICASSP.2015.7178788

DO - 10.1109/ICASSP.2015.7178788

M3 - Conference article in Journal

SP - 4330

EP - 4334

JO - I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings

JF - I E E E International Conference on Acoustics, Speech and Signal Processing. Proceedings

SN - 1520-6149

ER -