Adaptive threshold nonlinear algorithm for adaptive filters with robustness against impulse noise

A new adaptation algorithm for adaptive filters is proposed, introducing "adaptive threshold" in the nonlinear correlation function. The adaptive threshold varies according to the long-term average of the error signal plus the additive noise, which makes the adaptive filtering system highly robust against sharp increase in the noise yet maintains the fast convergence and low residual error properties. Results of simulation and theoretical calculation prove the effectiveness of the proposed algorithm and its superiority to the conventional LMS algorithm in an impulse noise environment.