An Improved Feed-Forward Neural Network Blind Equalization Algorithm Based on Construction Function

In QAM communication system, CMA without phase information was only utilized to module statistical property of signals. In the phase deviation channel, great phase error was brought out. Simultaneously, it affected the convergence rate. A restraint function utilizing the amplitude of the signal was constructed in this article. The function must approach zero when the amplitude was chosen. The cost function was converted into a restraint function. The restraint stem includes the information of module property and phase characteristic. A new neural network blind equalization based on construction function was realized. Computer simulation indicates that the algorithm overcomes QAM signal phase deviation, speeds up the convergence rate, and reduces bit error ratio.