Abstract—The paper provides the performance analysis of the optimal parameters selection for the weighted local polynomial approximation (LPA) estimators combined by a data-driven adaptive method used for the automatically adjusted nonparametric LPA estimator size detection. The provided examples show that the LPA estimators upgraded by the modified adaptive intersection of the confidence intervals (ICI) based method (called the RICI method) outperform those based on the original ICI rule, while the ICI rule based LPA estimators were known to outperform non-adaptive ones with fixed size window length. The method’s performance is analyzed in noise environment for two signals, showing the modified ICI based LPA estimators to be superior to those based on the original ICI method in terms of denoising estimation error reduction. The method can be applied in various technical fields, including image and video filtering, beamforming for phased array radar with antenna switching, acoustic echo cancellation, instantaneous frequency estimation, etc.

Index Terms—adaptive filtering, signal denoising, signal reconstruction, edge preserving

Cite: Jonatan Lerga, Marino Franušić, and Victor Sucic, "Parameters Analysis for the Time-Varying Automatically Adjusted LPA Based Estimators," Jounal of Automation and Control Engineering, Vol. 2, No. 3, pp. 203-208, September, 2014. doi: 10.12720/joace.2.3.203-208