The Periodic Signal Filtration Using the Robust Digital Filter Order Calculation Optimized by Approximation

Abstract

The article assesses the filtration of different shape periodic signals with optimized calculation of digital low pass filter (LPF) orders, i.e., the calculation based on minimizing the difference between approximated values and filtered values of a noisy signal. To assess the feasibility of the optimized digital filter order calculation, a series of experiments on generation and filtration of noisy periodic signals with different signal shapes and various signal-to-noise ratios has been conducted. For all of the noisy signals, optimal (relative to the original noise-free signals) and optimized (relative to approximated signals) LPF orders based on a minimum median absolute error (MAE) have been calculated. After calculation of MAE between filtered and original signals, it was observed that the optimal and optimized filters allow to get comparable noisy signal filtration quality. The article ends with conclusions about the possibility of using the optimized calculation of the digital filter order for close-to-optimal filtration (with the lowest MAE relative to the original noise-free signal) of noisy periodic digital signals.