Abstract:
In recent years, the demand for the reduction of impulsive noise (IN) has increased due to its enhanced effect on human health. In order to resolve this problem, number of techniques and adaptive algorithms are deployed in active noise control (ANC) system. The effective and most reliable method is the use of good adaptative algorithm for ANC system to reduce the IN.
The most commonly used FxLMS algorithm in ANC system diverges when encountered with IN. Therefore, in this research work, new adaptive algorithms are proposed to minimize the IN for ANC system. For this reason, first of all a new filtered-x least mean absolute third (FxLMAT) algorithm is tested in ANC for reduction of IN. The FxLMAT algorithm yields better convergence than the FxLMS algorithm and its variants but it also leads to instability in the presence of high IN. For further improvement in stability and convergence of FxLMAT algorithm, two modified versions of FxLMAT based on clipping and nulling are proposed i.e. sample ignored FxLMAT (SIFxLMAT) algorithm and sample clipped FxLMAT (SCFxLMAT) algorithm. The results depicted the better convergence and stability of proposed modification in the FxLMAT algorithm. These modifications are based on thresholding and become unstable too for cases of high impulses.
The limitations of proposed modifications in FxLMAT algorithms at high IN are overcome by proposing three effective new algorithms to achieve better convergence, stability and robustness. The proposed algorithms for ANC of IN sources are variable step-size filtered-x least mean absolute third (VSSFxLMAT) algorithm, filtered-x robust normalized least mean absolute third (FxRNLMAT) algorithm and hybrid variable step-size filtered-x robust normalized least mean absolute third (VSSFxRNLMAT) algorithm. The proposed VSSFxRNLMAT algorithm combines the step size equations of VSSFxLMAT and FxRNLMAT algorithms to amalgamate their benefits. Extensive computer simulations are performed, and outcomes verified the better stability, robustness and improved convergence of proposed hybrid VSSFxRNLMAT algorithm as compared to investigated and other proposed algorithms.
