The Comparative Study of Adaptive Channel Equalizer Based on Feed Forward Back Propagation, Radial Basis Function Neural Network(RBFNNs) & Least Mean Square (LMS) Algorithm

Artificial Neural networks (ANNs) have been extensively used in many signal processing applications. Linear & Nonlinear adaptive filters based on a variety of neural network models have been used successfully for system identification in a wide range of applications. Due to their capacity to form complex decision regions, ANNs have been most popularly applied, in particular, for channel equalization of digital communication channels. The mean square error (MSE) criterion, which is usually adopted in neural learning, is of interest in the channel equalization. In this paper, we introduce a novel approach to adaptive channel equalization using feed forward & Radial Basis Function neural network (RBFNNs) that exploits the principle of discriminative learning while minimizing error function. The performance of proposed method has been compared with gradient based algorithms such as LMS (Least Mean Square) which are often characterized by slow convergence. ANNs technique using fast learning feed forward configuration based on Back Propagation Algorithm, offers high speed of convergence w.r.t. LMS adaptive filtering algorithm. Computer simulation for the equalization of QAM signals in AWGN transmission channel is presented, which demonstrates the effectiveness of the proposed technique vis-à-vis LMS algorithm.