AUTHORS: Zouhour Ben Ahmed, Nabil Derbel
Download as PDF
ABSTRACT: Volterra models are very useful for representing nonlinear systems with vanishing memory. The main drawback of these models is their huge number of parameters to be estimated. In this paper, we present a new class of Volterra models, called Volterra-Parafac models, with a reduced parametric complexity, by considering Volterra kernels of order (p > 2) as symmetric tensors and by using a parallel factor (PARAFAC) decomposition. This paper is concerned with the problem of identification of third-order Volterra-PARAFAC models. Two types of algorithms are proposed for estimating the parameters of these models when input-output signals and kernel coefficients are real valued. The first is called Levenberg-Marquardt algorithm and the second is the Partial Update LMS algorithms. Some simulation results illustrate the proposed identification methods.
KEYWORDS: Volterra models, identification, tensors, PARAFAC, Levenberg-Marquardt, Partial Update LMS
REFERENCES:
[1] E. Mumolo and D. Francescato, Adaptive predictive coding of speech by means of Volterra predictors, In Proc. of IEEE Winter Workshop on Nonlinear digital signal processing, Tampere, Finland 1993
[2] Y. Kajikawa, Subband parallel cascade Volterra filter for linearization of loudspeaker systems, In EUSIPCO, Lausanne, Switzerland 2008
[3] L. Tan and J. Jiang, Adaptive Volterra filters for active control of nonlinear noise processes, IEEE Tr. on Signal Processing 49, 2001, pp. 1667–1676.
[4] A.Y. Kibangou and G. Favier, Blind equalization of nonlinear channels using tensor decompositions with code/space/time diversities, Signal Processing 89, 2009, pp. 133–143.
[5] C.A.R.. Fernandes and G. Favier and J.C.M. Mota , Blind identification of multiuser nonlinear channels using tensor decomposition and precoding, Signal Processing 89, 2009, pp. 2644–2656.
[6] G. Favier and T. Bouilloc, Identification de modle de Volterra base sur la decomposition PARAFAC, GRETSI, Dijon, France 2009
[7] G. Favier and A.Y. Kibangou and T. Bouilloc , Nonlinear system modeling and identification using Volterra-PARAFAC models, Int Journal on Adaptive Control and Signal Processing, 2011.
[8] A.Y. Kibangou, Modles de Volterra complexit rduite : Estimation paramtrique et application l’galisation des canaux de communication, Universit Nice-Sophia Antipolis-UFR Sciences, 2005.
[9] T. Bouilloc, Applications de dcompositions tensorielles l’identification de modles de Volterra et aux systmes de communication MIMO-CDMA, Universit Nice-Sophia Antipolis-UFR Sciences, 2011.
[10] D. Nion and L. De Lathauwer, Sparation et galisation aveugles de signaux CDMA par la dcomposition en blocs d’un tenseur au moyen de l’algorithme de Levenberg-Marquardt, 21th Colloque GRETSI, Troyes, France 2007
[11] G. Tomasi and R. Bro, A Comparison of Algorithms for Fitting the PARAFAC Model, Computational Statistics and Data Analysis 50, 2006, pp. 1700–1734.
[12] H. Gavin, The Levenberg-Marquardt method for nonlinear least squares curve-fitting problems, Dept. Civil and Environmental Engineering, Duke Univ 2011
[13] Z. Ben Ahmed and G. Favier and N. Derbel, Identification of Volterra-PARAFAC models using Partial Update LMS algorithms, 7th International Conference on Modelling, Identification and Control, Sousse, Tunisia 2015
[14] S.C. Douglas, Adaptive filters employing partial updates, IEEE Tr. on Circuits and Systems-II: Analog and digital signal processing 44, 1997, pp. 209–216.
[15] M. Godavarti and A.O. Hero, Partial Update LMS Algorithms, IEEE Tr. on Signal Processing 53, 2005, pp. 2382–2399.
[16] K. Dogancay and P.A. Naylor, Recent advances in partial update and sparse adaptive filters, in Proc. European Signal Processing Conference 2005
