Reweighted Zero-Attracting Modified Variable Step-Size Continuous Mixed p-Norm Algorithm for Identification of Sparse System Against Impulsive Noise

摘要

The efficiency of traditional adaptive filtering algorithms decreases for sparse system identification when compared to the identification of non-sparse systems. The modified variable step-size continuous mixed p-norm (MVSS-CMPN) algorithm for identifying non-sparse systems was developed in the presence of impulsive noise. In this work, a reweighted zero-attracting MVSS-CMPN algorithm is developed by inducing a sparse penalty function into the MVSS-CMPN algorithm to exploit the sparsity of the system under the effect of impulsive noise. From the simulations, it is found that the presented algorithm achieves a steady state of -25.16 dB for low sparsity in system identification scenario, whereas the MVSS-CMPN algorithm achieves a steady state of -9.96 dB.