Cluster-Sparse Proportionate NLMS Algorithm With the Hybrid Norm Constraint

In this paper, an enhanced proportionate normalized least mean square (PNLMS) algorithm with the hybrid l
_2,0
-norm constraint is proposed for block-sparse signal processing. The proposed algorithm penalizes a mixed l
_2,0
-norm on the PNLMS to fully exploit the sparsity and handle block-sparse signals, which is called the l
_2,0
norm constrained PNLMS (L20-PNLMS). The L20-PNLMS is well derived and carefully analyzed. Various experiments have been constructed to verify the effectiveness of the devised L20-PNLMS. The experimental results demonstrate that the devised L20-PNLMS performs better than the previous PNLMS algorithms do in block sparse signal processing.