A NEW LMS/NEWTON ALGORITHM FOR ROBUST ADAPTIVE FILTERING IN IMPULSIVE NOISE (WedAmOR5)

Author(s) : Yi Zhou (The University of Hong Kong, Hong Kong, P. R. China) S. C. Chan (The University of Hong Kong, Hong Kong, P. R. China) K. L. Ho (The University of Hong Kong, Hong Kong, P. R. China)

Abstract : This paper proposes a new LMS/Newton algorithm for robust adaptive filtering in impulse noise. The new algorithm is obtained by applying the non-linear filtering technique and the robust statistic approach to the conventional fast LMS/Newton method. A robust method for estimating the required threshold parameters for impulse suppression is also given. Simulation results show that aside from retaining the advantages of the LMS/Newton algorithm such as low complexity and numerical stability, the new algorithm is more robust and effective in suppressing the adverse effects of the impulses.

Menu