A Correcting Method for Pitch Extraction Using Neural Networks

Akio OGIHARA; Kunio FUKUNAGA

A Correcting Method for Pitch Extraction Using Neural Networks

Akio OGIHARA, Kunio FUKUNAGA

Full Text Views

0

Cite this

Summary :

Pitch frequency is a basic characteristic of human voice, and pitch extraction is one of the most important studies for speech recognition. This paper describes a simple but effective technique to obtain correct pitch frequency from candidates (pitch candidates) extracted by the short-range autocorrelation function. The correction is performed by a neural network in consideration of the time coutinuation that is realized by referring to pitch candidates at previous frames. Since the neural network is trained by the back-propagation algorithm with training data, it adapts to any speaker and obtains good correction without sensitive adjustment and tuning. The pitch extraction was performed for 3 male and 3 female announcers, and the proposed method improves the percentage of correct pitch from 58.65% to 89.19%.

Publication: IEICE TRANSACTIONS on Fundamentals Vol.E77-A No.6 pp.1015-1022

Publication Date: 1994/06/25

Publicized

Online ISSN

DOI

Type of Manuscript: Special Section PAPER (Special Section of Papers Selected from 1993 Joint Technical Conference on Circuits/Systems, Computers and Communications (JTC-CSCC'93))

Category: Neural Networks

Cite this

Copy

Akio OGIHARA, Kunio FUKUNAGA, "A Correcting Method for Pitch Extraction Using Neural Networks" in IEICE TRANSACTIONS on Fundamentals, vol. E77-A, no. 6, pp. 1015-1022, June 1994, doi: .
Abstract: Pitch frequency is a basic characteristic of human voice, and pitch extraction is one of the most important studies for speech recognition. This paper describes a simple but effective technique to obtain correct pitch frequency from candidates (pitch candidates) extracted by the short-range autocorrelation function. The correction is performed by a neural network in consideration of the time coutinuation that is realized by referring to pitch candidates at previous frames. Since the neural network is trained by the back-propagation algorithm with training data, it adapts to any speaker and obtains good correction without sensitive adjustment and tuning. The pitch extraction was performed for 3 male and 3 female announcers, and the proposed method improves the percentage of correct pitch from 58.65% to 89.19%.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/e77-a_6_1015/_p

Copy

@ARTICLE{e77-a_6_1015,
author={Akio OGIHARA, Kunio FUKUNAGA, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={A Correcting Method for Pitch Extraction Using Neural Networks},
year={1994},
volume={E77-A},
number={6},
pages={1015-1022},
abstract={Pitch frequency is a basic characteristic of human voice, and pitch extraction is one of the most important studies for speech recognition. This paper describes a simple but effective technique to obtain correct pitch frequency from candidates (pitch candidates) extracted by the short-range autocorrelation function. The correction is performed by a neural network in consideration of the time coutinuation that is realized by referring to pitch candidates at previous frames. Since the neural network is trained by the back-propagation algorithm with training data, it adapts to any speaker and obtains good correction without sensitive adjustment and tuning. The pitch extraction was performed for 3 male and 3 female announcers, and the proposed method improves the percentage of correct pitch from 58.65% to 89.19%.},
keywords={},
doi={},
ISSN={},
month={June},}

Copy

TY - JOUR
TI - A Correcting Method for Pitch Extraction Using Neural Networks
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 1015
EP - 1022
AU - Akio OGIHARA
AU - Kunio FUKUNAGA
PY - 1994
DO -
JO - IEICE TRANSACTIONS on Fundamentals
SN -
VL - E77-A
IS - 6
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - June 1994
AB - Pitch frequency is a basic characteristic of human voice, and pitch extraction is one of the most important studies for speech recognition. This paper describes a simple but effective technique to obtain correct pitch frequency from candidates (pitch candidates) extracted by the short-range autocorrelation function. The correction is performed by a neural network in consideration of the time coutinuation that is realized by referring to pitch candidates at previous frames. Since the neural network is trained by the back-propagation algorithm with training data, it adapts to any speaker and obtains good correction without sensitive adjustment and tuning. The pitch extraction was performed for 3 male and 3 female announcers, and the proposed method improves the percentage of correct pitch from 58.65% to 89.19%.
ER -