A generalized decision-directed blind equalization algorithm applied to equalization of multipath rayleigh fading mobile communication channels
J. Karaoguz, Sasan H. Ardalan · 1992
This dissertation presents a generalized decision-directed blind adaptive equalization algorithm based on ideas from neural networks. The algorithm converges to the desired solution for closed-eye channel situations for which existing blind equalization techniques fail to converge. Simulation results demonstrate the superior performance of the proposed algorithm, applied to the equalization of time-varying multipath fading mobile communication channels. A global understanding of blind equalization techniques is developed by establishing a relationship between the proposed algorithm and the existing blind equalization algorithms. It is shown that the generalized decision-directed algorithm provides an alternative approach in choosing the nonlinear estimator at the output of the equalizer. The sigmoid estimator, used by the generalized decision-directed algorithm, is shown to be the closest to the Bayes estimator, compared to the estimators used by the existing algorithms. The convergence behavior of the proposed blind equalization algorithm is described. The existence of the desired solution at the equilibrium is proved using the Bussgang property possessed by the algorithm. The existence of another suboptimal solution is also reported. It is shown that the generalized decision-directed algorithm actually converges to the desired solution, not to the suboptimal one, for an infinite tap equalizer. An efficient and simple implementation of the generalized decision-directed blind equalization algorithm is proposed using neuron structures and VLSI technology. The implementation has the desirable capability of providing a smooth transition from the Bayes estimator for high channel distortions, to the minimum MSE criterion for low channel distortions, during the equalization process.