A PNLMS algorithm with individual activation factors

Authors:
Francisco das Chagas de Souza;Orlando José Tobias;Rui Seara;Dennis R. Morgan
Affiliations:
LINSE, Circuits and Signal Processing Laboratory, Department of Electrical Engineering, Federal University of Santa Catarina, Florianópolis, Santa Catarina, Brazil;LINSE, Circuits and Signal Processing Laboratory, Department of Electrical Engineering, Federal University of Santa Catarina, Florianópolis, Santa Catarina, Brazil;LINSE, Circuits and Signal Processing Laboratory, Department of Electrical Engineering, Federal University of Santa Catarina, Florianópolis, Santa Catarina, Brazil;Bell Laboratories, Alcatel-Lucent, Murray Hill, NJ
Venue:
IEEE Transactions on Signal Processing
Year:
2010

Citing 7
Cited 3

Natural gradient works efficiently in learning

Neural Computation
Non-negative Matrix Factorization with Sparseness Constraints

The Journal of Machine Learning Research
Adaptive algorithms for sparse echo cancellation

Signal Processing
Acoustic MIMO Signal Processing (Signals and Communication Technology)

Acoustic MIMO Signal Processing (Signals and Communication Technology)
A low delay and fast converging improved proportionate algorithm for sparse system identification

EURASIP Journal on Audio, Speech, and Music Processing
Exploiting sparsity in adaptive filters

IEEE Transactions on Signal Processing
Proportionate adaptive algorithms for network echo cancellation

IEEE Transactions on Signal Processing

Stochastic model for the mean weight evolution of the IAF-PNLMS algorithm

IEEE Transactions on Signal Processing
Fast communication: Two variants of the sign subband adaptive filter with improved convergence rate

Signal Processing
On the stochastic modeling of the IAF-PNLMS algorithm for complex and real correlated Gaussian input data

Signal Processing

Quantified Score

Hi-index	35.69

Visualization

Abstract

This paper presents a proportionate normalized least-mean-square (PNLMS) algorithm using individual activation factors for each adaptive filter coefficient, instead of a global activation factor as in the standard PNLMS algorithm. The proposed individual activation factors, determined in terms of the corresponding adaptive filter coefficients, are recursively updated. This approach leads to a better distribution of the adaptation energy over the filter coefficients than the standard PNLMS does. Thereby, for impulse responses exhibiting high sparseness, the proposed algorithm achieves faster convergence, outperforming both the PNLMS and improved PNLMS (IPNLMS) algorithms.