A Comparison of Non-Intrusive SNR Estimation Algorithms and the Use of Mapping Functions
File(s)
Accepted version
Author(s)
Eaton, D
Brookes, DM
Naylor, PA
Type
Conference Paper
Abstract
We present a comparative evaluation of six methods for non-intrusive Signal-to-Noise Ratio (SNR) estimation for narrowband speech in noise. We demonstrate that the performance of all methods can be improved by applying a non-linear mapping function to their estimates of SNR. We have employed phrases built from the TIMIT speech corpus and noises from a broad range of sources including ITU-T P.501, NOISEX-92, and Soundjay. We compare the accuracy of the methods in estimating the SNR of both stationary and non-stationary noise and we conclude that with the mapping function, the best current methods can estimate the SNR to within approximately 3.5 dB for SNRs from -5 dB to 35 dB.
Date Issued
2013-09
Date Acceptance
2013-03-15
Citation
Proc. European Signal Processing Conf. (EUSIPCO), 2013, pp.1-5
ISBN
978-0-9928626-0-2
Publisher
EURASIP
Start Page
1
End Page
5
Journal / Book Title
Proc. European Signal Processing Conf. (EUSIPCO)
Copyright Statement
© 2013 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works.
Sponsor
Home Office
HMGCC
Engineering and Physical Sciences Research Council
Grant Number
PO 7073101
P/O: A169643
1097854
Source
EUSIPCO
Subjects
speech enhancement
SNR
Noise estimation
Publication Status
Published
Start Date
2013-09-09
Coverage Spatial
Marrakech, Morocco