A Simulation Investigation of Modified FxLMS Algorithms for Feedforward Active Noise Control

In this paper, two modified FxLMS algorithms are proposed based on the post-masking-based LMS (PMLMS) algorithm. They are the PMl-FxLMS and the signed PMl-FxLMS (SPMl-FxLMS) algorithms. In both algorithms, l denotes the length of the error signal memory. The control filter coefficients are updated by the maximum absolute value in the error signal memory, instead of the immediate value. The difference between the two modified FxLMS algorithms is that the PMl-FxLMS algorithm keeps the sign of the error sample with the largest absolute value, while the SPMl-FxLMS algorithm uses the sign of the immediate error sample. The simulation results show that the SPMl-FxLMS algorithm converges faster than the standard FxLMS algorithm with the same step-size, and the PMl-FxLMS algorithm may be difficult to converge when l is large.