Adaptive Algorithms Against Impulsive Interference And Their Performance Analysis

The research on adaptive filtering algorithms under impulsive noise environment is an important topic in the area of signal processing.The sign-error least mean square(SE-LMS)algorithm,which is developed from minimizing the cost function of the absolute error,can effectively resist the interference on the adaptive filter,which is caused by the impulsive interference.However,on one hand,if the input signal for the adaptive filter is corrupted when estimating an unknown system,the convergence rate of the SE-LMS algorithm will be slow,and the steady-state misalignment will increase.On the other hand,distributed estimation finds widespread applications in many areas,but the SE-LMS algorithm cannot obtain good convergence performance when it is directly used in distributed estimation.In this thesis,the idea of unbiased criterion is firstly applied to the SE-LMS algorithm,and a bias-compensated SE-LMS(BC-SE-LMS)algorithm is proposed.This algorithm cannot only compensate the bias introduced by the input noise but also shows relatively strong robustness against impulsive noise.Besides,this thesis also analyzes the mean and mean-square performances to provide theoretical guide for the use of this algorithm in practical applications.Secondly,this thesis analyzes the mean-square steady-state and stability performance of the diffusion sign-error least mean square(DSE-LMS)algorithm,which can improve the performance analysis theory of the algorithm.Finally,this thesis applies the idea of dual sign to the distributed network,and proposes a diffusion dual SE-LMS(DDSE-LMS)algorithm,and theoretical mean and mean square performance analysis of the algorithm is carried out.This algorithm can be employed in distributed estimation under impulsive noise environment.As compared to the DSE-LMS algorithm,the proposed DDSE-LMS algorithm has a fast convergence speed or a low steady-state misalignment.