A Robust Algorithm for Derivation of Heart Rate Variability Spectra from ECG and PPG Signals

Abstract

Digital spectral analysis of Heart Rate Variability (HRV) signals provides quantitative markers of the Autonomic Nervous System (ANS) activity, underpinning a variety of physiologic processes. In this investigation, a robust algorithm was developed to derive HRV signals, their FFT- or AR-based spectra and their standard spectral features from either electrocardiogram (ECG) or photoplethysmographic (PPG) signals. ECG/PPG signals were first interpolated to a common sampling rate and detrended using a first order DC-notch IIR high-pass filter to remove very low frequencies. Next, the undecimated Discrete Wavelet Transform (DWT) Daubechies-6 family of filters was used to selectively remove some of the high-frequency subbands from the signals. The filtered ECG/PPG signals were then squared to increase the dynamic range of the target dominant peaks from which accurate peak-to-peak intervals could be extracted. A smart peak-tracking and monitoring algorithm was used to detect the peaks while enforcing a valid trend of the instantaneous heart rate. Once the peak-to-peak intervals were obtained, the HRV signal was determined using cubic spline interpolation to create a signal at a very low sampling rate of 1-4 Hz. Standard Power Spectrum Estimation (PSD) of the HRV signal was then performed to generate normalized LF, HF and LF/HF spectral features to quantify parasympathetic influences and sympathovagal balance.