Correntropy induced metric penalized NLMF algorithm to improve sparse system identification

Adaptive sparse system identification (ASIDE) techniques have been successfully applied in many applications, such as sparse channel estimation and radar target detection. By using ℓ1-norm sparse constraint, two sparse normalized least mean fourth (NLMF)-type algorithms, i.e., zero-attracting NLMF (ZA-NLMF) and reweighted ZA-NLMF (RzA-NLMF) have been developed. The two algorithms can achieve fast tracking speed and good steady-state performance. However, there still exists a performance gap between the two developed algorithms and lower bound due to the fact that either ZA-NLMF or RZA-NLMF does not fully exploit system sparsity. This paper proposes an improved sparse NLMF algorithm by using correntropy induced metric which can take advantage of system sparsity more efficient. Computer simulations are given to confirm the validity of the proposed algorithm.