Performance Comparison of Modified Non-Linear FxLMS Algorithm for Impulsive Noise Based on ANCs

This paper presents versions of adaptive algorithm for active control of noise sources that are of impulsive nature. In the recent literature, statistical modeling of impulsive noise is usually done with the help of Alpha Stable Distribution. For stable distribution, the variance or second order moment is infinite. The standard adaptive algorithms like FxLMS algorithm which usually works on the principle of minimizing the finite variance becomes non-convergent or divergent in case of stable distributions because FxLMS algorithm is designed on the basis of reducing error by following the principle of Mean Square Error. To improve the performance of FxLMS algorithm in Active Noise Control Systems for impulsive noise sources, a modification of standard FxLMS algorithm has been proposed in this paper. This modification is done with the help of non-linear functions namely sigmoid, hyperbolic tangent and logistic sigmoid individually in order to improve the convergence of error for FxLMS algorithm when the noise source is impulsive. Convergence performance is measured for each modified version and the level of impulsiveness is also varied in each case in order to determine which version is the best error converging option. Computational complexity is also calculated for each modified version in order to analyze the performance of each algorithm in a multidimensional fashion.