LMS algorithm for active noise control with improved gradient estimate

Most existed LMS algorithms have noisy gradient estimate since they use instantaneous square error to estimate ensemble average of the square errors. In this paper, a new type of LMS algorithm is developed for active noise control. The gradient estimate in the proposed algorithm is improved by using the weighted time average instead of the instantaneous value to approximate the ensemble average. Compared to the filtered-x LMS algorithm, the proposed algorithm is shown to have many improved convergence properties. Computer simulations are given to support the theoretical analysis.