A RBF equalizer using fast clustering algorithm

A RBF (radial basis function) network has been used in digital communication channel equalization for its identical structure to the optimal Bayesian symbol by symbol decision equalizer. The most important problem in the equalization using RBF is the fast searching of the correct centers (channel states). The k-means clustering algorithm has been used to find the desired channel states. In this paper, the problem can be transformed into finding the representative centers suitable for channel states using the interrelation among the elements of the center set. Computer simulation results show its fast convergence speed and less training iteration than equalization using the traditional k-means clustering algorithm.