Open AccessCompensator models based on block-oriented neural networks
Author(s) -
Elena Solovyeva
Publication year - 2019
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1333/3/032080
Subject(s) - cascade , artificial neural network , computer science , control theory (sociology) , block (permutation group theory) , distortion (music) , set (abstract data type) , volterra series , algorithm , nonlinear system , mathematics , artificial intelligence , engineering , bandwidth (computing) , amplifier , telecommunications , physics , geometry , control (management) , quantum mechanics , chemical engineering , programming language
Block-oriented neural networks represented as a cascade structure of linear dynamic circuits and inertialess nonlinearities are used for the non-linear compensator synthesis. The behavioral models of these neural networks describe the mapping of the set of input signals into the set of output signals. A block-oriented architecture is convenient for building the compensator structure inverse to the structure of the distorting system. Under the stated approach, the compensator model is simpler than universal models, for instance, the Volterra series. This fact is demonstrated when modelling a non-linear compensator based on the neural Hammerstein network. This compensator is applied for the suppression of non-linear distortion in the digital communication channel described by the Wiener model.