Stationarity

A spectrum describes a recording as a mix of steady waves: so much power at the breathing rhythm, so much at the 10-second blood-pressure rhythm, and so on. That description only makes sense if the mix stays the same for the whole recording. Statisticians call such a series stationary.

Real recordings break the rule easily. Standing up halfway, starting to talk, a cough, or falling asleep changes heart rate and its rhythms, and the spectrum then averages two different states into one that never existed. The 1996 Task Force standards ask for short-term recordings made under stable conditions for this reason.

Why it matters for HRV.

• Keep conditions fixed. Laborde, Mosley and Thayer (2017) ask for no movement during a recording and a settled posture before it starts. That is what keeps a 5-minute reading stationary.
• Split, do not smooth. Detrending Detrending Detrending Removing the slow drift from a series of beat intervals before a spectrum is calculated, so the drift does not pile up as false power in the lowest frequencies. More about Detrending removes a slow drift, but it cannot fix a sudden change. The diagram below shows a step of 90 ms halfway through a recording: detrending removes most of it but leaves a jolt where the change happened, which the spectrum then counts as rhythm. The fix is to analyse the two halves separately.
• Do not compare lengths. A longer recording takes in more change, so its measures are not comparable with a short one’s.

A step the spectrum cannot handle

The teaching recording, altered

Gap between beats, resampled at 4 Hz ±130 ms around the mean; dashed is the trend

After detrending ±130 ms around zero

0 min 5 min

Two line charts of the gap between beats over five minutes. The top one drops by about 90 ms halfway through and stays lower; the bottom one is level again around zero, with a short jolt where the step was and the breathing wave on top.

See also detrending and how to take a good HRV reading.