I'm trying to understand the following formula in this paper:

enter image description here

Sadly, the paper does not define most of the symbols it is using, so my guesses so far are:

  • $T_{tr}$ refers to the sampling interval
    • i.e. $\frac{1}{44100} s$
  • $\phi$ refers to phase
    • specifically: $\phi(k_{tr}, (n-1)h)$ refers to the angle of the complex vector representing frequency $k_{tr}$ for the FFT from time point $(n-1)h$ to $nh$
  • so basically this is saying: out of all FFT frequency bins, only include those frequencies $k_{tr}$ where above equation holds true

I tried googling around to see if there's a standard definition of "transient frequency bins" (like there is for FFT frequency bins) this is referencing, but no success. I also don't understand

  • Why would one compare radians $\phi$ with a unit of time $T_{tr}$?
    • Is this maybe a stand-in for 0 to account for sampling errors?
  • Why isn't $T_{tr}$ just called $T$ as it seems to be a property of the signal and unrelated to the transient bins?
    • (If this is not the sampling interval, then how would one obtain $T_{tr}$?)

How does this formula define $K_{tr}$?


1 Answer 1


without reading the paper,

Eqn 4 has the form of a central difference form of the second derivative. It is a numerical second derivative compared to a threshold. It differs from the standard form in that it centers on $n-1$.


Whatever $\phi$ is, it is the positive change compared to a threshold.

  • $\begingroup$ well the second derivative of phase is the rate of change of frequency. the sweep rate. but if these are phases of a frequency component at adjacent sample times, this sweep rate (and second derivative) is radians per sample per sample.and cannot be compared dimensionally to $\frac{1}{44100}$ second. $\endgroup$ Jul 31, 2018 at 5:00

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.