Feedback active noise control based on forward-backward LMS predictor

Pavithra , S and Narasimhan, SV (2012) Feedback active noise control based on forward-backward LMS predictor. Signal, Image and Video Processing. pp. 1-9. ISSN 1863-1711

PDF FBANC_FBLMS_JOUR_Feb.pdf Restricted to Registered users only Download (256kB) | Request a copy

Abstract

In this paper, a new feedback active noise control (FBANC) system based on Forward-Backward error LMS (FBLMS) predictor, is proposed. The misadjustment of the FBLMS predictor is about half that of the forward error LMS (FLMS) predictor. The new ANC system employs FBLMS predictors both for its main path (MP) predictor and for the noise canceller (NC) for the secondary path (SP) identification (SPI). To realize the MP predictor based on the FBLMS concept, a new FXLMS structure is proposed. But for the NC for the SPI, the FBLMS predictor is directly used. The MP predictor based on FBLMS reduces its misadjustment. Further the use of FBLMS predictor for the NC, as it gives a good prediction of primary noise component in the error (residual noise), improves the SNR for SPI. Thus the improved SP estimate and the reduced misadjustment for the MP predictor achieved, result in a significantly better overall noise reduction (of about 8dB) over the ANC that uses the MP predictor and noise canceller for SPI, both based only on the forward error LMS algorithm. The computational load for the proposed algorithm is about twice that of FBANC that uses only forward error.

Item Type:	Article
Additional Information:	Copyright to this article belongs to /s. Springer-Verlag
Uncontrolled Keywords:	Forward–backward LMS predictor;Transform-domain forward–backward LMS (TFBLMS) predictor;FXLMS algorithm based on TFBLMS predictor;Feedback active noise control
Subjects:	ENGINEERING > Communications and Radar ENGINEERING > Electronics and Electrical Engineering
Depositing User:	Ms. Alphones Mary
Date Deposited:	03 Jan 2013 06:10
Last Modified:	03 Jan 2013 06:10
URI:	http://nal-ir.nal.res.in/id/eprint/11410

Actions (login required)

View Item