Fig. 1

The flowchart of the proposed approach to study the long-term recording HRV. The time series shown in the top is constructed by the RR interval via the interpolation, and smoothed for the visualization purpose. (1) The truncated segments of the heart rate time series are marked in red. In this step, we apply the concentration of frequency and time, which is composed of the synchrosqueezing transform and the generalized multitaper algorithm, to estimate the power spectrum associated with different segment. To enhance the visualization, the power spectrum is plotted in the log2 scale. (2) All estimated power spectra are stitched together according to time, which ends up with an image (time–frequency representation) called the time-varying power spectrum generated by ConceFT. (3) Determine the dominant curve in the tvPS and evaluate the energy around the dominant curve at each time. Then evaluate the ratio of the energy concentrated on the band marked by the red dashed curves and the remaining energy. As a result, we obtain the non-rhythmic to rhythmic ratio (NRR) function. (4) The standard deviation of the NRR function is the summary index called the NRR index. This index quantifies how concentrated the tvPS is around the dominant spectral component