A Study on Long Term Physiological Index by Using Time-Frequency Analysis

Abstract

In this thesis eight physiological indices: heart rate (HR) reflection index (RI) stiffness index (SI) standard deviation of normal-to-normal interval (SDNN) root mean square of successive differences (RMSSD) very low frequency (VLF) low frequency(LF) and high frequency (HF) are recorded for long term analysis These indices are computed from photoplethysmograph signals and used to develop a model for the level of fatigue By visualizing the signals I observed that the modes of high frequencies are affected by the modes of low-frequencies which are related to breath In order to make the signals more stable I use the empirical mode decomposition (EMD) of Hilbert-Huang transform (HHT) to remove the modes of breathing After that I use two different time-frequency analysis methods short-time Fourier transform (STFT) and synchrosqueezing wavelet transform (SST) to compute the HR SDNN and RMSSD I have recorded physiological signals for 323 times in the four months and then analyse these records by using different methods logistic regression decision tree support vector machine nearest neighbors and naive Bayers By logistic regression I make a personal predictive model of fatigue so that I can judge whether I am tired or not In addition by Fourier analysis I found that from the spectrum of HR RI SI SDNN and RMSSD they show a major cycle 7 days Besides from the spectrums of VLF LF and HF a major cycle every 28 days can be found Keyword: physiological indices heart rate stiffness index reflection index Hilbert-Huang transform short time Fourier transform synchrosqueezing wavelet transform

Date of Award	2017 Feb 16
Original language	English
Supervisor	Yu-Chen Shu (Supervisor)