Blind Radial Basis Function Network Equalizer for Digital Communication Channels

The design of adaptive equalizers is an important topic for practical implementation of efficient digital communications. The application of a radial basis function neural network (RBF) for blind channel equalization is analyzed. The proposed architecture shows the design process of a radial basis function equalizer, in which the number of basis function used, is substantially fewer than conventionally required. The reduction of centers is accomplished in two steps. First an algorithm is used to select a reduced set of centers, which lies close to the decision boundary. Then the centers in this reduced set are grouped and an average position is chosen to represent each group. Channel order and delay, which are determining the factors in setting the initial number of centers, are estimated from regression analysis. This center reduction can be done by simple sorting operation, which corresponds to the weight initialization. Finally the weight is adjusted iteratively by an unsupervised least mean square (LMS) algorithm. Since the process of weight initialization using the under lying structure of the RBF equalizer is very effective, the proposed blind RBF equalizer can achieve almost identical performance with MMSE equalizer. The resulting structure is modular and real-time implementation is feasible using simple hardware. The validity of proposed equalizer is demonstrated by computer stimulation.