An Improved A Priori MMSE Spectral Subtraction Method for Speech Enhancement

Author

Cheng, Ning ; Liu, Wen-Ju ; Xu, Bo

Author_Institution

Chinese Acad. of Sci, Beijing

fYear

2007

fDate

23-27 Sept. 2007

Firstpage

373

Lastpage

377

Abstract

This paper presents a modification of conventional minimum mean squared error (MMSE) spectral subtraction algorithm by introducing an adaptive averaging factor to accurately estimate the a priori SNR. Performance of the modified a priori SNR is evaluated by comparing with conventional spectral subtraction algorithm. Improved results are obtained in terms of speech quality measures for various types of noise when the time-frequency varying averaging factor, proposed in this paper, is adapted in the conventional subtraction rules.

Keywords

least mean squares methods; speech enhancement; time-frequency analysis; SNR; a priori MMSE spectral subtraction method; adaptive averaging factor; minimum mean squared error; speech enhancement; time-frequency varying averaging factor; Additive noise; Frequency estimation; Laboratories; Noise level; Noise measurement; Noise reduction; Pattern recognition; Signal to noise ratio; Speech enhancement; Time frequency analysis; MMSE spectral subtraction; a priori SNR; adaptive averaging factor; speech enhancement;

fLanguage

English

Publisher

ieee

Conference_Titel

Signal Design and Its Applications in Communications, 2007. IWSDA 2007. 3rd International Workshop on

Conference_Location

Chengdu

Print_ISBN

978-1-4244-1074-3

Electronic_ISBN

978-1-4244-1074-3

Type

conf

DOI

10.1109/IWSDA.2007.4408401

Filename

4408401