Spectral Mapping Using Kernel Principal Components Regression for Voice Conversion

The Gaussian mixture model (GMM) method is popular and efficient for voice conversion (VC), but it is often subject to overfitting. In this paper, the principal component regression (PCR) method is adopted for the spectral mapping between source speech and target speech, and the numbers of principal...

Full description

Saved in:
Bibliographic Details
Main Authors: Peng SONG, Li ZHAO, Yongqiang BAO
Format: Article
Language:English
Published: Institute of Fundamental Technological Research Polish Academy of Sciences 2013-03-01
Series:Archives of Acoustics
Subjects:
Online Access:https://acoustics.ippt.pan.pl/index.php/aa/article/view/5
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850104170443440128
author Peng SONG
Li ZHAO
Yongqiang BAO
author_facet Peng SONG
Li ZHAO
Yongqiang BAO
author_sort Peng SONG
collection DOAJ
description The Gaussian mixture model (GMM) method is popular and efficient for voice conversion (VC), but it is often subject to overfitting. In this paper, the principal component regression (PCR) method is adopted for the spectral mapping between source speech and target speech, and the numbers of principal components are adjusted properly to prevent the overfitting. Then, in order to better model the nonlinear relationships between the source speech and target speech, the kernel principal component regression (KPCR) method is also proposed. Moreover, a KPCR combined with GMM method is further proposed to improve the accuracy of conversion. In addition, the discontinuity and oversmoothing problems of the traditional GMM method are also addressed. On the one hand, in order to solve the discontinuity problem, the adaptive median filter is adopted to smooth the posterior probabilities. On the other hand, the two mixture components with higher posterior probabilities for each frame are chosen for VC to reduce the oversmoothing problem. Finally, the objective and subjective experiments are carried out, and the results demonstrate that the proposed approach shows greatly better performance than the GMM method. In the objective tests, the proposed method shows lower cepstral distances and higher identification rates than the GMM method. While in the subjective tests, the proposed method obtains higher scores of preference and perceptual quality.
format Article
id doaj-art-e74a284ba7824012a94c7c7bf48a3fd5
institution DOAJ
issn 0137-5075
2300-262X
language English
publishDate 2013-03-01
publisher Institute of Fundamental Technological Research Polish Academy of Sciences
record_format Article
series Archives of Acoustics
spelling doaj-art-e74a284ba7824012a94c7c7bf48a3fd52025-08-20T02:39:23ZengInstitute of Fundamental Technological Research Polish Academy of SciencesArchives of Acoustics0137-50752300-262X2013-03-01381Spectral Mapping Using Kernel Principal Components Regression for Voice ConversionPeng SONG0Li ZHAO1Yongqiang BAO2Key Laboratory of Underwater Acoustic Signal Processing of Ministry of Education Southeast UniversityKey Laboratory of Underwater Acoustic Signal Processing of Ministry of Education Southeast UniversitySchool of Communication Engineering, Nanjing Institute of TechnologyThe Gaussian mixture model (GMM) method is popular and efficient for voice conversion (VC), but it is often subject to overfitting. In this paper, the principal component regression (PCR) method is adopted for the spectral mapping between source speech and target speech, and the numbers of principal components are adjusted properly to prevent the overfitting. Then, in order to better model the nonlinear relationships between the source speech and target speech, the kernel principal component regression (KPCR) method is also proposed. Moreover, a KPCR combined with GMM method is further proposed to improve the accuracy of conversion. In addition, the discontinuity and oversmoothing problems of the traditional GMM method are also addressed. On the one hand, in order to solve the discontinuity problem, the adaptive median filter is adopted to smooth the posterior probabilities. On the other hand, the two mixture components with higher posterior probabilities for each frame are chosen for VC to reduce the oversmoothing problem. Finally, the objective and subjective experiments are carried out, and the results demonstrate that the proposed approach shows greatly better performance than the GMM method. In the objective tests, the proposed method shows lower cepstral distances and higher identification rates than the GMM method. While in the subjective tests, the proposed method obtains higher scores of preference and perceptual quality.https://acoustics.ippt.pan.pl/index.php/aa/article/view/5spectral mappingoverfittingoversmoothingdiscontinuitykernel principal component regression
spellingShingle Peng SONG
Li ZHAO
Yongqiang BAO
Spectral Mapping Using Kernel Principal Components Regression for Voice Conversion
Archives of Acoustics
spectral mapping
overfitting
oversmoothing
discontinuity
kernel principal component regression
title Spectral Mapping Using Kernel Principal Components Regression for Voice Conversion
title_full Spectral Mapping Using Kernel Principal Components Regression for Voice Conversion
title_fullStr Spectral Mapping Using Kernel Principal Components Regression for Voice Conversion
title_full_unstemmed Spectral Mapping Using Kernel Principal Components Regression for Voice Conversion
title_short Spectral Mapping Using Kernel Principal Components Regression for Voice Conversion
title_sort spectral mapping using kernel principal components regression for voice conversion
topic spectral mapping
overfitting
oversmoothing
discontinuity
kernel principal component regression
url https://acoustics.ippt.pan.pl/index.php/aa/article/view/5
work_keys_str_mv AT pengsong spectralmappingusingkernelprincipalcomponentsregressionforvoiceconversion
AT lizhao spectralmappingusingkernelprincipalcomponentsregressionforvoiceconversion
AT yongqiangbao spectralmappingusingkernelprincipalcomponentsregressionforvoiceconversion