16
$\begingroup$

I need to conduct some soundscape analysis collected in bird colonies on remote islets. Seabirds are loud! I end up with files with regular audio saturation/clipping. The waveform is pretty obvious. We reduced the gain but even though the birds move around and sometimes get very close and saturate the mic anyway.

My question is: which type of analyses are affected by audio saturation and which one are not and could be acceptable ?enter image description here

$\endgroup$
1
  • 1
    $\begingroup$ Could you please explain in more detail what soundscape analysis entails. This will help targeted answers with relation to whether the analyses are/not affected by clipping. Additionally, a spectrogram of the same audio clip may provide more background if available. $\endgroup$
    – Thejasvi
    Jul 4, 2022 at 13:21

2 Answers 2

5
$\begingroup$

Not at all, be only carful on the results you present. For example, the frequency modulation should still be visible in your data. You only cannot determine the power level (only a lower limit). However, these type of signals are good for replica based filtering (i.e. classification by means of matched filter, cross correlation, etc). IMHO, clipped signals are much more useful than low SNR signals.

$\endgroup$
3
$\begingroup$

I don't think there's any value to relying on acoustic measurements from segments of the WAV file that are clipped. You can't trust any spectral info on a clipped segment, for example.

However, you can still work with the segments that are unclipped. The rest of this answer will depend on what your goals are with your soundscape analysis. If you are considering seabirds are part of the soundscape (this is appropriate, but depends on your project's goals), then run your third octave level (TOL) analysis or create your power spectral densities (PSDs) as you normally would, but only on the non-clipped segments.

$\endgroup$
1
  • 1
    $\begingroup$ To add to this, I would also remove a buffer around your clipped segments. For example, if you have a bird that is just sitting on the mic and screaming, and your file is clipping irregularly because of that, I would remove the entire time the bird is calling, not just the clipped parts because you are not able to capture the full range of the bird's vocalizations. Including partial calls can skew your TOL analysis. $\endgroup$ Jul 5, 2022 at 6:50

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

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