search

orcasound / ambient-sound-analysis

This repository aims to hold code for UW MSDS capstone project analyzing ambient sounds in orcasound hydrophone data
MIT License
4 stars 4 forks source link

Test if passage of a ship produces pattern as expected #37

Closed valentina-s closed 5 months ago

valentina-s commented 7 months ago

Timestamp for a passing boat (I believe):

@scottveirs is this the pattern you are looking for?

scottveirs commented 7 months ago

Looking at this test plot more carefully than I was able to do during the standup this morning, @valentina-s, it doesn't look totally as expected for "typical" Orcasound spectrogram including a vessel noise event. Two reasons:

  1. Usually ocean spectrograms have higher power at low frequencies than at higher frequencies (see e.g. this Dosits page or search for Wentz curves). It's possible there was some sort of high-frequency noise source (e.g. electrical contamination or a rain event?), but in any case the overall spectrogram is sort of "reversed" from typical.
  2. While I still think (as I did this morning) that the event centered on 04:09:41 and ~500Hz is akin to a vessel noise signature (the smiling shapes are common and known as Lloyd's mirror effect, see screenshot below from this [paper](https://www.researchgate.net/publication/288664513_Communication_masking_in_marine_mammals_A_review_and_research_strategy/figures?lo=1
  3. )), the duration is way to short for the passing of a distant ship; lasting only 20 seconds, it's more likely a very fast speedboat at close range (and those passages usually look more tonal with horizontal bands rather than the smiling shapes.
Screenshot 2024-01-17 at 1 41 59 PM
scottveirs commented 7 months ago

An even more compelling test plot (other than a spectrogram) IMHO would be a plot of broadband level vs time over a few hours. For example, a plot like this (without the red AIS-based range estimates) might be optimal:

Screenshot 2024-01-17 at 1 44 49 PM

Playing around with smoothing was a key part of the methods presented by Val in the corresponding ASA 2018 talk on "delta metrics".

ben-hendricks commented 7 months ago

OLAB_PTTD_CSTM-01_sample

veirs commented 7 months ago

@vaibhavmehrotraml @zprice12 @ttan06 I am sorry I missed your Friday Meet to talk about checking psd codes. I was all bollixed up with frozen pipes and cars sliding off the road and other weather worries. Now, all is calm again.

I want to join with you three and Valentina and Ben and Scott and to do so, I wonder if you guys have selected a particular audio data file to use for inter-comparisons. If you have such a wav file or prefer to specify a particular minute and a particular node, please let me know and I will run my 'legacy' codes to compare with your modern ones. By 'legacy' I mean codes that I have used for a long time and that don't use many 'fancy' new methods from scipy or labrosa etc.

zprice12 commented 5 months ago

Ship passage now produces expected pattern in the code