Nonlinear Channel Equalization Using Hermit Artificial Neural Network
Paper ID : 1485-IST
1Amin Naemi *, 2mohammadmehdi Homayounpour
1Amirkabir University of Technology
2Amirkabir university of technology
Channel equalization of modulated signals may be defined as a classification problem where the number of classes determines the type of modulation. Neural networks are employed for channel equalization. Because of nonlinear processing in neural networks, these networks can approximate nonlinear decision boundaries. In this paper, we propose a computationally efficient neural network for equalization of nonlinear telecommunications channels for signals with 4-QAM modulation, based on Hermit neural network (HeNN). This equalizer has one layer and classify samples by expanding the input pattern by Hermitian polynomials. The performance of the proposed equalizer is compared with LMS, MLP, RBF, and FLANN equalizers through extensive computer simulations. The results show that the proposed equalizer provides better results in terms of MSE, training time, constellation diagram and accuracy.