I'm currently working with physiological signals (PPG and GSR) for emotion recognition but, from my research, I've found out that almost everyone in that area use a PSD analysis over a FFT analysis. I've been reading about them and found out that PSD helps with giving a clearer view of the spectrum despite the amount of data that you have, as described in this blog https://blog.endaq.com/why-the-power-spectral-density-psd-is-the-gold-standard-of-vibration-analysis, also because it is supposed to be used with random signals and therefore GSR and PPG are signals that have some random nature within them. Despite this, I still can't gasp the intuition why a PSD is used over a FFT analysis.

  • 1
    $\begingroup$ The PSD is a property of a stochastic signal. The FFT is a piece in a few methods to estimate that property. Your question is equivalent to "why do people measure the length of things instead of using a ruler?". $\endgroup$ Sep 3 '20 at 20:33
  • 1
    $\begingroup$ Most algorithms that I know of that estimate the PSD of a measured digital signal use fast Fourier transforms under the hood. You could do it differently, but there's a reason that the first "F" in "FFT" stands for "fast". $\endgroup$
    – TimWescott
    Sep 3 '20 at 20:57