I have quite a few noisy signals and I want to calculate their spectral tilt over time, preferably using a method from literature. So far, I can only come up with the slope of the line between the F0 and F1 peaks. The problem is, I don't know how to calculate the spectral tilt by hand for the below example (which is a cross-cut of the energy-frequency spectrum), because F1 is so low.

I am not using a spectrogram here, but a cochleogram, but it should work the same.

The frequencies and energies of all the peaks are the following:

 Freq (Hz) Energy (dB)
    95.465 -46.700
   142.06  -43.072
   178.24  -43.564
   269.82  -14.422
   405.93  -34.140
   537.26  -19.539
   803.47  - 2.081
  1060.3   -10.409
  1315.0   -14.574
  1581.0   -24.363
  1843.0   -29.801
  2214.9   -34.464
  2744.3   -28.979
  3197.9   -26.913
  3961.1   -30.493

Can someone help me with this?

  • 1
    $\begingroup$ i dunno how the lit defines "spectral tilt" but if there is an agreement regarding how amplitude and frequency are scaled, i imagine it's the slope of some sorta weighted least-square fit for a straight line. if you decide that it's log-amplitude vs. log-frequency and you specify a frequency range for the data, you can compute a value in dB per octave or dB per decade. do you know how to do least-squares fit to data? $\endgroup$ Jul 30 '14 at 18:16
  • $\begingroup$ You're right. I should first wonder what spectral tilt actually is. I'll come back if I have that clear. $\endgroup$
    – Lewistrick
    Jul 31 '14 at 13:56

Your Answer

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

Browse other questions tagged or ask your own question.