Abstract:

To improve the performance of the least mean square (LMS) algorithm, an adaptive LMS algorithm of Krylov subspace based on convex combination is proposed. In this approach, the Krylov subspace transform is firstly performed to obtain the sparse structure of the unknown system impulse response in the Krylov subspace domain, and then an improved proportionate normalized LMS (IPNLMS) algorithm and a variable taplength normalized LMS (VTNLMS) algorithm are combined. Finally, the opposite Krylov subspace transform are performed to obtain unknown system impulse response. Simulation results show both the fast convergence rate and the small steady state mean square deviation (MSD) of the proposed algorithm.