A novel convex combination of LMS adaptive filter for system identification

The least mean square (LMS) algorithm is most popular adaptive filter because of its low-cost and robust. However, its convergence rate is slow when the measurement noise is added in unknown system. The combination of two least mean square (CLMS) filters is developed to address the tradeoff in many signal processing applications. Based on the analysis of basic-CLMS algorithm, a novel sign adaptation scheme for convex combination of adaptive filters is proposed with instantaneous transfer scheme. By using an adaptive sign adaptation scheme, the proposed scheme slightly reduces the computational complexity of the basic combination of mixing parameter, also improves the robustness of the mixing parameter. And, by employing instantaneous transfer scheme, the proposed algorithm benefits from the fast convergence rate during the period of convergence transition. The simulation studies in the context of system identification show that the proposed algorithm with instantaneous transfer scheme has lower computational complexity and faster convergence rate than that of the basic-CLMS algorithm during the period of convergence transition.