Robust estimators for speech enhancement in real environments

Yuma Sandoval-Ibarra, Victor H. Diaz-Ramirez, Vitaly Kober

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Common statistical estimators for speech enhancement rely on several assumptions about stationarity of speech signals and noise. These assumptions may not always valid in real-life due to nonstationary characteristics of speech and noise processes. We propose new estimators based on existing estimators by incorporation of computation of rank-order statistics. The proposed estimators are better adapted to non-stationary characteristics of speech signals and noise processes. Through computer simulations we show that the proposed estimators yield a better performance in terms of objective metrics than that of known estimators when speech signals are contaminated with airport, babble, restaurant, and train-station noise.

Original languageEnglish
Title of host publicationOptics and Photonics for Information Processing IX
EditorsMohammad A. Matin, Abdul A. S. Awwal, Andres Marquez, Mireya Garcia Vazquez, Khan M. Iftekharuddin, Mireya Garcia Vazquez, Khan M. Iftekharuddin, Abdul A. S. Awwal, Mohammad A. Matin, Andres Marquez
PublisherSPIE
ISBN (Electronic)9781628417647, 9781628417647
DOIs
StatePublished - 2015
Event9th Conference of Optics and Photonics for Information Processing - San Diego, United States
Duration: 10 Aug 201512 Aug 2015

Publication series

NameProceedings of SPIE - The International Society for Optical Engineering
Volume9598
ISSN (Print)0277-786X
ISSN (Electronic)1996-756X

Conference

Conference9th Conference of Optics and Photonics for Information Processing
Country/TerritoryUnited States
CitySan Diego
Period10/08/1512/08/15

Keywords

  • Robust estimators
  • rank-order statistics
  • speech enhancement
  • statistical estimators

Fingerprint

Dive into the research topics of 'Robust estimators for speech enhancement in real environments'. Together they form a unique fingerprint.

Cite this