Need help quantifying high frequency peaks

Hi, I’m doing some research on coral reefs and passive acoustics and I am attempting to quantify the number of high frequency peaks in a sample. I am assuming each high frequency peak is a snap made by a snapping shrimp and thus I should be able to predict reef quality by quantifying the number of high frequency peaks.

Specifically in the image I’ve attached, I’m looking for a way in which I can count the number of straight vertical lines that appear (which are the shrimp snaps).

I am on Audacity 2.3.3 and Windows 10.

Are all of the red vertical lines in that screenshot sounds that you want to count? (including the darker red lines to the right of those that you have labelled).

What is the time scale of that screenshot?

Could you post a short sample, that includes a couple of the “snaps”, in WAV format. Just a few seconds.

As well as removing clicks, Paul-L’s De-Clicker plugin will isolate clicks.
Then you could truncate the silence between the isolated clicks,
The duration of the truncated track, relative to the original audio, would be a measure of its clickiness.

quantifying clickiness.gif

First off I just want to thank both of you for replying so quickly and so intelligently. I’m pretty much entirely new to using Audacity and doing anything sound related so this help is very greatly appreciated.

@steve I’ve attached a very small clip of one of the samples I was messing around with. It might be hard to hear in that short clip but the vacuumy sound is a sea scooter and the sound of bacon frying should be the snaps of the snapping shrimp.

@trebor That is an extremely helpful plugin and if it works, exactly what I’m looking for. Thank you so much.

@Trebor: Using the plugin that you have linked me leads to some potentially promising results for my first sample. Using the exact numbers allowed me to come to a similar result where I was able to pretty much cut out all the unnecessary whirring and reduce the sample to a fraction of its original size. However, because I’m exceedingly inexperienced in Audacity and other sound programs, I am not positive that I used the optimal values for the De-Clicker and the truncating silence effects. I may have cut out some clicks unknowingly, which would affect the data.

All I’m really certain is that snapping shrimp snap at frequencies of about 2 kHz - 300 kHz so I’ve just used the high pass filter to filter out any sound below 2 kHz. I’ve attached a screenshot of what I managed to scrape together for my first sample and the second sample (which is much longer and contains magnitudes more clicks).

@Steve: Sorry I forgot to answer the question, but yes, I would be counting/quantifying each red vertical line.

Your microphone / hydrophone appears to have a frequency response up to about 16 kHz. (You can adjust the vertical scale to see this region:

Below about 5 kHz, there’s a lot of other sound, making it difficult to pick out the “snaps”. So we need to concentrate on the frequency band 5 to 16 kHz.

We don’t need to do anything with frequencies above 16 kHz because there’s virtually nothing there.

I’d suggest mixing the track down to mono, so that audio in both channels gets treated the same (“Tracks menu > Mix > Mix Stereo down to Mono”)

Here’s the track again after applying a very steep high-pass filter using the “Equalization” effect at around 6 kHz, with about 14 dB gain in the pass-band:

And now, amplified with the “Amplify” effect (default settings) in normal “Waveform” view:

I presume that the bigger clicks are shrimps that are close to the microphone, and faint clicks are shrimp further away. Does that sound reasonable?
If so, then we need to decide what level to consider as “within range”. We are limited by the noise floor, so in this case I’d suggest a noise floor setting of around -18 dB (You will need to experiment with this figure to find the optimum for your recordings)
(If you are testing multiple audio files, and you want to compare one file with the other, then you will need to use exactly the same Amplify and “noise floor” settings for each file).

So now, I’ll use “Sound Finder” ( with these settings:

And this is the result:
(It’s not perfect, so you need to estimate the amount of error).


Having seen Steve’s method, it looks more scientific than mine. (I suspect it’s quicker to apply too).