Robust Sparse EHCAF Algorithm and Convex Combination Based Variants

This paper presents a novel approach for enhancing the performance of adaptive filtering in environments with non-Gaussian or impulsive noise by developing a sparsity-aware exponential hyperbolic cosine adaptive filter (EHCAF) algorithm to exploit the sparsity of the unknown system to improve identification accuracy under impulsive noise conditions. The convergence behaviour of the proposed algorithm is analyzed statistically. The impact of step size and zero-attracting coefficients on convergence speed and steady-state mean square deviation (MSD) shows a trade-off that is tackled with a convex combination (CC) of two sparsity-aware EHCAF adaptive filters. Performance validation is conducted through simulations, evidencing the effectiveness of our method in achieving superior filtering outcomes for sparse systems in challenging noise conditions.