A Genetic Approach for Linear-Quadratic Channel Identification With Usual Communication Inputs

The blind identification of a special class of nonlinear system is pursued in this paper. In particular a genetic algorithm is developed for the blind identification of linear-quadratic Volterra model excited by inputs commonly used in digital communication such as PSK and QAM signals. Since the cost function with higher order statistics has local minimum points, the use of genetic algorithm allows to escape from these last and to find an optimal solution of the identified channel. Several simulations are performed and show a fair accuracy given sufficiently long observation records.