An improved feedback neural network for the design of all-pass phase equalizers

An improved neural-based approach for the design of FIR all-pass phase equalizer with prescribed magnitude and phase responses is introduced. The error differences in the frequency domain are formulated as a Lyapunov energy function. By mapping the objection function to the corresponding Hopfield neural network, the optimal filter coefficients are therefore obtained using a parallel manner. Simulation results indicate that the proposed technique achieves good performance as compared to existing methods.