Abstract
In this paper, the tracking behavior of the epsilon-normalized sign-error least mean square (NSLMS) algorithm is analyzed in the presence of white and correlated Gaussian regressors. Moreover, generic analytical expressions are derived for the optimal step-size and the corresponding optimal mean-square error (MSE) of the epsilon-NSLMS algorithm for both the real-and complex-valued data cases. Additionally, a comparison between the convergence behavior of the epsilon-NSLMS algorithm and the epsilon-normalized least mean square (NLMS) algorithm is also discussed. Finally, simulation results to corroborate our theoretical findings are presented.