DocumentCode
2707126
Title
Performance of Voice Conversion Systems Based on GMM and Applied to Arabic Language
Author
Guerid, A. ; Houacine, A.
Author_Institution
Commun.´´s Lab. & Signals Treatment, USTHB, Algiers, Algeria
fYear
2009
fDate
28-30 Dec. 2009
Firstpage
118
Lastpage
121
Abstract
Voice conversion (VC) consists in modification of the source interlocutor’s voice toward another voice of the target interlocutor. In this article, we are studying the different performances of the standard Arabic speech conversion and in particular that of the vocal tract. We have used a regularized discrete cepstrum in order to estimate the spectrum of the speech signal and the parameters of a GMM model. We have raised the number or amount of the mixture Gaussians to 512 in order to calculate the conversion function.
Keywords
Cepstrum; Covariance matrix; Frequency; Gaussian distribution; Gaussian processes; Informatics; Laboratories; Probability; Speech; Virtual colonoscopy; Gaussien Mixture Model (GMM); Voice conversion;
fLanguage
English
Publisher
ieee
Conference_Titel
MEMS, NANO, and Smart Systems (ICMENS), 2009 Fifth International Conference on
Conference_Location
Dubai, United Arab Emirates
Print_ISBN
978-0-7695-3938-6
Electronic_ISBN
978-1-4244-5616-1
Type
conf
DOI
10.1109/ICMENS.2009.19
Filename
5489359
Link To Document