A Variable Step Size Leaky Least Mean Square Adaptive Algorithm

Abstract

This article proposes a variable step-size leaky least mean square (VS-LLMS) adaptive algorithm for an adaptive filter. The proposed variable step-size algorithm is based on the criterion of squared error autocorrelation accumulation. In addition, the proposed VS-LLMS algorithm hascomputational complexity lower than the VSS-LLMS algorithm. Under the system identification model in stationary process, the theoretical analysis which is the steady state behavior of the algorithm is also proposed in the paper. The simulation results have been shown that, they have trend corresponding with those of the analytical results. Furthermore, the performance comparison of the proposed adaptive algorithm by simulation is also compared to the other algorithms such as LLMS and VSS-LLMS algorithms under the same testing condition of the system. The results have shown that the VS-LLMS algorithm have low level steady-state mean square deviation (MSD) whose valuesare lower than other algorithms. Therefore, the proposed algorithm outperforms the LLMS and VSS-LLMS algorithms.