A Sigmoid Filter Proportionate Normalized LMS Algorithm for Sparse Systems

Formerly, a least mean square algorithm employing a sigmoid framework (SLMS) was developed to cope with the impulsive noise. However, the SLMS technique was sparse agnostic, and the LMS approach has a gradient noise amplification issue. To address the point, the SLMS algorithm is updated to sigmoid normalized LMS (SNLMS), and to adjust for sparsity, a filter proportionate adaptation principle is applied to the SNLMS method, yielding the SFPNLMS algorithm. Finally, simulation results show that the innovative SFPNLMS technique is more resistant in an impulsive environment and has a rapid convergence for identification of the sparse systems than other current algorithms.