A Variable Step Size of the LMS Algorithm for System Identification

The Least Mean Square (LMS) algorithm and adaptive filter techniques have been used in system identification. However, their effectiveness may degrade due to the learning gain in the LMS algorithm is a constant. In this paper, a variable step-size learning parameter (VSSP) is proposed to improve the performance of the unknown system identification. The proposed algorithm is implemented without any offline learning phase, while faster convergence can be achieved. Moreover, the computational complexities of different methods are compared. Comparative simulation results demonstrate the validity of the proposed method.