155 0

Delayless individual-weighting-factors sign subband adaptive filter with band-dependent variable step-sizes

Title
Delayless individual-weighting-factors sign subband adaptive filter with band-dependent variable step-sizes
Author
남상원
Keywords
Band-dependent variable step-size; delayless structure; individual weighting factor; l(1)-norm minimization; sign subband adaptive filter
Issue Date
2017-07
Publisher
IEEE-INST ELECTRICAL ELECTRONICS ENGINEERS INC
Citation
IEEE-ACM TRANSACTIONS ON AUDIO SPEECH AND LANGUAGE PROCESSING, v. 25, no. 7, page. 1526-1534
Abstract
Sign subband adaptive filter algorithms with individual weighting factors (IWF-SSAF) were recently proposed to achieve improved convergence performance in impulsive noise environments by fully utilizing the decorrelating property of subband adaptive filters. However, such sign subband adaptive filter (SSAF) algorithms may have an inherent signal path delay problem for real-time applications, and thus, can be restrictively applied. In this paper, a delayless IWF-SSAF with band-dependent variable step-sizes (BD-VSS), robust in impulsive noise environments, is proposed for real-time applications, whereby two delayless filter structures developed for the l(2) -norm-based SAF are employed along with the IWF-SSAF. In particular, a BD-VSS algorithm is also introduced for better convergence by applying the recent l(1) norm minimization technique to respective subband. Finally, the performance of the proposed delayless IWF-SSAF with BD-VSS is verified in various impulsive interference environments such as system identification with impulsive noise, acoustic echo cancellation in a double-talk scenario, and active impulsive noise control.
URI
https://ieeexplore.ieee.org/document/7914653https://repository.hanyang.ac.kr/handle/20.500.11754/114917
ISSN
2329-9290
DOI
10.1109/TASLP.2017.2699325
Appears in Collections:
COLLEGE OF ENGINEERING[S](공과대학) > ELECTRONIC ENGINEERING(융합전자공학부) > Articles
Files in This Item:
There are no files associated with this item.
Export
RIS (EndNote)
XLS (Excel)
XML


qrcode

Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.

BROWSE