Coefficient Pruning for Higher-Order Diagonals of Volterra Filters Representing Wiener-Hammerstein Models

Zeller M, Kellermann W (2008)

Publication Language: English

Publication Type: Conference contribution

Publication year: 2008

Conference Proceedings Title: Proc. Int. Workshop on Acoustic Echo and Noise Control (IWAENC)

Event location: Seattle (WA) US

URI: http://www.iwaenc.org/proceedings/2008/start.html

Abstract

Block-based nonlinear structures as, e.g., the Wiener-Hammer- stein model, are popular for analyzing a broad class of nonlinear distortions. On the other hand, these models can also be related to Volterra filters in diagonal coordinates which represent very general nonlinear filters with memory. In this work, we propose an approach for estimating the significant coefficients of the nonlinear Volterra kernels during adaptation of such block-based filters by relating the linear impulse response and the diagonals of higher-order kernels. As the number of coefficients for the higher-order kernels is gener- ally very large, this information can then be used to prune the num- ber of necessary coefficients and thus lower the computational com- plexity. Experimental results for several test signals and nonlinear systems demonstrate the effectiveness of such an approach in a non- linear acoustic echo cancellation scenario.

Authors with CRIS profile

Marcus Zeller Lehrstuhl für Multimediakommunikation und Signalverarbeitung Walter Kellermann Professur für Signalverarbeitung

How to cite

APA:

Zeller, M., & Kellermann, W. (2008). Coefficient Pruning for Higher-Order Diagonals of Volterra Filters Representing Wiener-Hammerstein Models. In Proc. Int. Workshop on Acoustic Echo and Noise Control (IWAENC). Seattle (WA), US.

MLA:

Zeller, Marcus, and Walter Kellermann. "Coefficient Pruning for Higher-Order Diagonals of Volterra Filters Representing Wiener-Hammerstein Models." Proceedings of the Int. Workshop on Acoustic Echo and Noise Control (IWAENC), Seattle (WA) 2008.

BibTeX: Download