Title :
Volterra kernels extraction from neural networks for amplifier behavioral modeling
Author :
Misic, Jelena ; Markovic, Vera ; Marinkovic, Zlatica
Author_Institution :
Fac. of Electron. Eng., Univ. of Nis, Nis, Serbia
Abstract :
In wireless communication systems the amplifier non-linear distortion problems are extremely challenging. The linearization techniques based on behavioral models of amplifiers, seem to be very promising, therefore developing a suitable non-linear model is of the crucial importance. A rigorous approach for non-linear modeling is using the of Volterra series, however the calculation of the Volterra coefficients is a complex and time-consuming task. In this paper, an easy and advanced approach for extraction of the Volterra kernels will be presented. The third order Volterra kernels are derived from the parameters of a feed-forward time delay neural network with a suitable activation function.
Keywords :
Volterra series; amplifiers; feedforward neural nets; linearisation techniques; nonlinear distortion; amplifier behavioral modeling; amplifier nonlinear distortion problems; feedforward time delay neural network; linearization techniques; neural networks; volterra kernels extraction; wireless communication systems; Artificial neural networks; Biological neural networks; Kernel; Mathematical model; Neurons; Solid modeling; Volterra kernels; Volterra series model; artifical neural networks; non-linear system modeling;
Conference_Titel :
Telecommunications (BIHTEL), 2014 X International Symposium on
Conference_Location :
Sarajevo
Print_ISBN :
978-1-4799-8038-3
DOI :
10.1109/BIHTEL.2014.6987646