Abstract :
The recently introduced normalised robust mixed-norm (NRMN) algorithm improved the performance of its predecessor robust mixed-norm (RMN), proposed for system identification under impulsive noise (IN) environments. The NRMN inherited the mixing parameter of RMN, the definition of which however was not based upon any statistical information for the IN. An improved performance NRMN algorithm is proposed which defines its mixing parameter and normalised step-size in a novel way based upon a model choice for the IN. The proposed algorithm outperforms the NRMN with respect to the convergence rate and misadjustment.
