A Bayesian Estimator for Non-intrusive Speech Quality Evaluation in Psychoacoustic Domain

Author

Chen, Guo ; Parsa, Vijay

Author_Institution

Dept. of Electr. & Comput. Eng., Western Ontario Univ., Ottawa, Ont.

fYear

2006

fDate

Aug. 2006

Firstpage

438

Lastpage

441

Abstract

A Bayesian estimator for non-intrusive speech quality evaluation is presented. In this novel speech quality estimator, the Gaussian mixture density hidden Markov models were used for characterizing different speech quality categories and the speech quality estimation was performed by using the Bayesian inference and minimum mean-squared error estimation. The performance of the proposed estimator was demonstrated by experimental evaluations in comparison with the standard ITU-T P.563 using speech coded databases

Keywords

Bayes methods; Gaussian processes; hidden Markov models; mean square error methods; speech processing; Bayesian estimator; Bayesian inference; Gaussian mixture density hidden Markov models; minimum mean-squared error estimation; nonintrusive speech quality; psychoacoustic domain; Bayesian methods; Distortion measurement; Feature extraction; Hidden Markov models; Psychology; Speech analysis; Speech coding; Speech enhancement; Speech processing; Testing; Bayesian inference; Hidden Markov model; Pitch power density; Speech quality evaluation;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Processing and Information Technology, 2006 IEEE International Symposium on

Conference_Location

Vancouver, BC

Print_ISBN

0-7803-9753-3

Electronic_ISBN

0-7803-9754-1

Type

conf

DOI

10.1109/ISSPIT.2006.270841

Filename

4042283