A New Partial-Update NLMS Adaptive-Filtering Algorithm

A partial-update NLMS (PU-NLMS) algorithm is proposed that uses a variable step size which is obtained by solving a constrained minimization problem. The proposed algorithm can be used with two different known updates of the inherent diagonal matrix. Simulation results in a system identification application demonstrate that the proposed PU-NLMS algorithm yields reduced steady-state misalignment as compared to the known PU-NLMS, the set-membership PU-NLMS, and the M-max NLMS algorithms. The proposed PU-NLMS algorithm requires approximately the same number of iterations to converge as the conventional and set-membership PU-NLMS algorithms and somewhat fewer iterations relative to the M-max NLMS algorithm. Furthermore, it is shown that through the use of one of the two known updates of the inherent diagonal matrix, reduced computational effort can also be achieved relative to those of the known PU-NLMS and M-max NLMS algorithms.