R. Pandey
Blind equalization, recurrent neural network, constant modulusalgorithm, decision feedback equalization
The nonlinear structure of a neural network-based blind equalizer yields a significant improvement in performance in comparison to equalizers with linear transversal filter structures. However, to reduce the burden of the feedforward neural equalizers, recurrent neural networks (RNN) are preferred for equalization. In this work two different complex valued recurrent neural structures— a fully connected RNN and a decision feedback equalizer (DFE) like multilayer perceptron (MLP)-based recurrent structure—are used for blind equalization. Their learning rules are derived using constant modulus algorithm (CMA). The activation functions used in the nodes of the networks are complex valued and suitable for quadrature amplitude modulation (QAM) signals. The performance of the MLP-based recurrent equalizer is found to be better than the fully connected RNN blind equalizer on the basis of mean square error (MSE) and symbol error rate (SER).
Important Links:
Go Back
