Adaptive Equalization of Nonlinear Time Varying-Channels using Radial Basis Network

The paper investigates adaptive equalization of nonlinear time varying digital communication channel. An architecture of equalization was proposed based on the Bayesian theory (R. Assaf et al., 2005) where an implementation by radial basis function neural network (RBFNN) was accomplished. We treated the equalization of binary transmission signal through dispersive nonlinear time varying channel. The hybrid training algorithm is used. For the supervised part, it uses the sequential LMS algorithm which has a good convergence over batch LMS algorithm. For the unsupervised part, the rival penalized competitive learning method is used, with the LBG algorithm for the initial values. The performance of the equalizer is compared with the Bayesian equalizer which has the optimal parameters