2
$\begingroup$

I have sperm whales recordings and I would like to isolate clicks and their bottom and surface reflections.

I first thought to run the echo detector associated with the click detector module but the help page is empty and I have found no information on its functioning and parametrisation online. At the same time, I found some (unclear for biologist) informations for the echoDetector packages, which seems to be able to perform what I would like to do.

My questions are:
What is the relation between echo detector of the click detector module and the echoDetector packages? (Is the second implemented in the first?)
If not, is echoDetector available for users (I don't find it in the Add modules... menu)?
More generally how to use click detectors in PAMGuard?

If you have or know where to find some clear and concise documentation on the subject I would also be interested.

Improve this question
$\endgroup$

1 Answer 1

Reset to default
2
$\begingroup$

A google search gave me

"EchoDataUnit: This data unit is for measuring multipath echo delay It requires raw acoustic data and can provide an estimate of the multipath delay computed via cepstral analysis."

rant: This unit will ONLY deal with ECHOs, if the recorder is attached to the whale. In typical PAM applications you are observing REFLECTIONS and not ECHOs. ECHOs are reflections that come back to the whale! \rant

Apart from that, the key word is cepstral analysis. This takes the whole time series (how much depends on the expected delay) and looks for repeating features. I let the explanation of the PAMGuard implementation to more active users, but I expect that there should be an input for the maximum delay.

Note, that, IIRC, PAMGuard uses cepstrum analysis also for sperm whale Inter-Pulse-Interval (IPI) estimation, that is a similar problem (head internal reflection) but with much less data (some 10s of ms).

Concerning, estimation of surface/bottom reflection, cepstral analysis may not work for time extended reflections due to rough sea and bottom interface.

Edit: to address the comment below this answer, the way I would it is to run a click/transient detector and to classify detection manually (e.g. using excel) as direct arrival, IPI, surface/bottom reflection.

Improve this answer
$\endgroup$
4
  • $\begingroup$ Thank you for your answer, If I had to summarise it, would you agree with "trying to use an echo detector (no matter which) will probably not work since echoes are not reflections"? I also read (a little) about the IPI detector and was wondering if it could tweak it to serve my purpose, I appreciate that you answered the question. If I'm not misunderstanding your answer, I think that the most reliable way to solve my original problem would then to simply run the click train detector and semi-manually associate click trains to a first arrival/bottom/surface reflection's value $\endgroup$
    – Patou
    Commented Mar 9 at 17:13
  • $\begingroup$ Yes, semi-manually associating the clicks is probably how I would start too. In my experience, I have found that making a time-series plot of estimated mult-path can be quite effective for this -- especially if used with a "data brush." I wrote a bit about this in my PhD thesis (Chapter 4 section 4.4): ourarchive.otago.ac.nz/handle/10523/2173?show=full. @WMXZ:Merriam-Webster dictionary defines echo as "the repetition of a sound caused by reflection of sound waves," without any specific mention of the need to return to the producer of the sound. $\endgroup$
    – Brian Miller ♦
    Commented Mar 13 at 0:42
  • $\begingroup$ key is "return", that IMO means to sender. (e.g. mail returns to sender, not to another person) $\endgroup$
    – WMXZ
    Commented Mar 13 at 6:13
  • $\begingroup$ @WMXZ, I hear you, but my point was that the definition of echo in English dictionaries does not contain the word "return." so I have no issues if people want to use "echo" as a synonym for multipath reflection. $\endgroup$
    – Brian Miller ♦
    Commented Mar 15 at 1:07

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.