期刊名称：Journal of Theoretical and Applied Information Technology
印刷版ISSN：1992-8645
电子版ISSN：1817-3195
出版年度：2017
卷号：95
期号：22
页码：6056
出版社：Journal of Theoretical and Applied
摘要：In this paper, the error-Gaussian-kernelled input of the algorithm developed by maximization of zero-error probability of constant modulus error (MZEP-CME) is studied for developing a method to reduce the weight perturbation of the MZEP-CME under impulsive noise. The proposed method is to normalize the input of MZEP-CME with the norm of the error-Gaussian-kernelled input (EGKI) in order to reduce weight perturbation. Then the denominator of the step size can make the algorithm unstable when it has a very small value or wide fluctuations. To prevent these incidents, a balanced power of EGKI between the current power and the past one is employed. This normalization with balance power provides an additional function for reducing further the weight perturbation in impulsive noise environment. Simulation results show that the weight fluctuation after convergence of the proposed algorithm is below half of that of the MZEP-CME. Also compared with the MZEP-CME, the proposed approach lowers the steady state MSE (mean squared error) by about 1 dB under impulsive noise.
