Channel equalization using neural networks

The equalization of different communication channels with different signaling constellations using artificial neural networks is investigated. We show that applying a fuzzy rule to the adjustment of the learning rate and momentum of the backpropagation network increases the convergence rate of the equalizer. We use the complex backpropagation network to equalize complex-valued constellations. Using the geometrical interpretation of the equalization problem, we propose a decision device which decides on whether the channel must be equalized by a linear equalizer or a neural network equalizer