Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
EVALUATION OF NOISE ROBUST FEATURES ON THE AURORA DATABASES Xiaodong Cui, Markus Iseli, Qifeng Zhu, and Abeer Alwan
 

Summary: EVALUATION OF NOISE ROBUST FEATURES ON THE AURORA DATABASES
Xiaodong Cui, Markus Iseli, Qifeng Zhu, and Abeer Alwan
Department of Electrical Engineering
University of California, Los Angeles, CA 90095
Email: xdcui, iseli, qifeng, alwan @icsl.ucla.edu
ABSTRACT
In this paper, we evaluate our noise robust feature extraction al-
gorithms on the Aurora 2 and the German part of Aurora 3. Sev-
eral algorithms are introduced and evaluated to deal with the noisy
speech signals including our previous noise robust techniques used
with Aurora 2, and new approaches evaluated with Aurora 3. Since
there exist some differences between the two databases, modifi-
cations of front-end modules are needed. For Aurora 2, the av-
erage error rate reduction is 47% for clean training and 12% for
multicondition training compared with the new baseline with end-
point detection. In Aurora 3, we obtain 17%, 27% and 53% er-
ror rate reduction for the well-matched, medium-mismatched and
high-mismatched cases, respectively.
1. INTRODUCTION
Aurora 2 [1] and Aurora 3 [2] are two databases used to test differ-

  

Source: Alwan, Abeer - Electrical Engineering Department, University of California at Los Angeles

 

Collections: Computer Technologies and Information Sciences