Automatic Bio-impedance Signal Analysis: Smoothing Processes Efficacy Evaluation in Determining the Vascular Tone Type

Automatic detection of feature points in bio-impedance signals is an essential step in signal analysis. In some cases, this process requires high degree of accuracy to ensure that the detected feature points result from changing bio-impedance not from inappropriate filter use or noise. To determine vascular tone by electrical bio-impedance, there are three basic feature points to be detected. These points are the first systolic wave peak, the diastolic wave peak, and the incisura. In this work, signal smoothing has been considered using three methods: Savitzky-Golay, Locally Weighted Polynomial Regression Method (LOESS), and Binomial smoothing. Each method has been applied to a specific signal which was recorded from thigh part of human body. The resulting signals have been considered in terms of different parameters. These parameters include clear appearance of feature points, change in feature points time, which results from smoothing process, and number of detected cycles that contain feature points after smoothing. The obtained results have been discussed according to each applied method. An evaluation criterion has been proposed to evaluate the applied smoothing methods. Based on the obtained results and discussions, it is clarified that each method can give better results according to the specified research goal.