Modifications of 4253HT Smoother in Extracting Signal from Noise

Abstract

The smoothing technique has been used broadly in data analysis to obtain sharp signals or patterns by eliminating noise or formless patterns from a data set. Various smoothers with great properties were reported in previous studies, mainly non-linear smoothers such as 4253HT smoother. Hence, the 4253HT was selected to be used in this study, with some modifications done on mean 2 with different types of Hanning. The modifications on mean 2 were done by substituting with Geometric, Quadratic, Harmonic, and Contraharmonic means. Meanwhile, three types of Hanning were used in this study, including Tukey, Shitan, and Husain’s type. Moreover, through a simulation study, this study aims to determine the best combination of smoother between four signals, which were Doppler, HeaviSine, Bumps, and Block signals. A Residual Mean Square Error (RMSE) was used as an evaluator to determine and assess the performance of modified 4253HT smoother when utilizing each smoother and noise level. As a result, Hanning Husain exhibited the best performance among all. Moreover, Hanning Tukey performed better on signal Block at lower noise levels (10% and 25%), while Hanning Shitan showed the worst performance. Besides, modified 4253HT using contra harmonic mean smoother recorded the most satisfying outcome in smoothing compared to other smoothers. Therefore, the 4253HT smoother will demonstrate the best performance with the combination of S5 smoother and Hanning Husain. Thus, it is suggested to utilize this smoother in further analysis, mainly in forecasting to provide accurate values and pattern for prediction.