Robust competitive diffusion LMS algorithm

The main drawback of the diffusion least mean square algorithm (DLMS) is its performance will degrade when the impulsive noise (IN) occurs in the system. To improve the performance, a robust competitive DLMS (RCDLMS) algorithm based on two types of effective IN detection method is proposed. Because both methods estimate noise variance to determine whether impulse noise occurs or not. The first method is using the shrinkage denoising method to obtain the variance of noise, and the second method is by estimating cross-correlation of the input and error signals to acquire the variance of noise. Finally, simulation results demonstrate that the proposed algorithm has a good tracking capability and robust performance against impulsive noise.