Gauss–Newton iteration method for radial basis function-based postdistortions in visible light communication systems

Abstract

In visible light communication systems, the inherent nonlinear characteristics of light-emitting diodes (LEDs) are the primary source of signal distortion. To alleviate the nonlinear effect of the LED and improve the system reliability, the polynomial method is used to construct the postdistorter. With strong processing ability and high solution accuracy, radial basis function (RBF) interpolation can also be used to alleviate the nonlinear effect. In terms of the iterative implementations, recursive least squares (RLS) algorithm has been widely adopted, but performance loss exists with the solution since RLS is initially proposed for linear regressions. Aiming at the nonlinear characteristics of post distortions, Gauss–Newton (GN) method is investigated to accomplish the iterations in this paper, where Taylor series expansion is used to approximate the nonlinear regression model. Simulation results show that the RBF can obtain better bit error ratio performance than polynomial, and the proposed GN iterations have significantly better performance compared to RLS in mitigating the nonlinear effect of LED.