Essential Feature Extraction of Photoplethysmography Signal of Men and Women in Their 20s

Abstract

This study aims to extract features of Photoplethysmography (PPG) signal of men and women in healthy subjects using Power Spectral Density (PSD) and Detrended Fluctuation Analysis (DFA). A PPG instrument was used to obtain the PPG signal of 15 men and 15 women. Using PSD, 4 of frequency bands were selected to divide the spectral component. The areas within frequency bands relative to the total area were computed as features of the signals. Furthermore, using DFA, the average fluctuation F(w) was computed. The feature extraction using this technique produces 4 features from different windows. Hurst exponent was calculated to analyse the characteristics of the time series. Additionally, F and T test for both techniques were computed to determine the differences between men and women features that have been gathered using these two techniques. The results indicated that the features of PPG signal of men and women using PSD and DFA are significantly different. In order to evaluate the results, a clustering analysis was applied to the results data using K-means clustering technique. The clustering graph shows that the features are well distributed into two groups.