Gun shot analysis shouldn't be much different from at least the initial step in analyzing seismic data, i.e. the distinction between
no event and the
initial pulse of an earthquake. The primary indicator is the rise-time of the event.
In the above example provided by
@KSTR you can easily see the sharp rises (i.e. short rise-times) at the on-set of each event (explosion of fireworks).
A very detailed explantation of seismic data analysis can be found here:
https://gfzpublic.gfz-potsdam.de/rest/items/item_4009_4/component/file_4010/content?download=true
the rise-time that you talk about, actually is what I thought to be a problem, because if there are two (incredibly) rapid events, and the first one is quieter, I don't have a reliable way to tell apart if the first event is part of the rapid onset, or a different one, I'll have to research to see how others have dealt with this problem (in the earthquake case it's definitely easier because it's differentiating between no earthquake VS earthquake, it doesn't have the problem of wondering if it was a mini earthquake lasting a few milliseconds followed by the big earthquake event, if that makes sense)
now for the most interesting part:
And to give a perspective on what has been investigated in the field of gunshot detection, have a look at 2019 paper "Development of Computational Methods for the Audio Analysis of Gunshots" by Ryan Lilien.
i'm not sure how I've never come across this, I'll have to read it in more detail but I really appreciate that there has already been research on this, there seem to be lots of methods / algorithms tested, so this is a treasure chest of information... also appreciate the website providing all of the test data (but I think if I were to go the neural network route, I think the best results would be achieved with training with my pre-labeled data, since lots of things affect it, even the microphone type)
-------------
Now for a new question - which hasn't been tackled here before is the possibility of utilizing for example
TWO (or three) microphones - would this make the job easier? (sorry for the long read)
because... as the following quote says - I'd also want to achieve the direction...
Also, you might be wandering into the military use for telling where shots are coming from. The systems they use hear the sound and then can point you in the exact direction of the sound. I'm not sure without looking into it in depth, I think they can even tell the distance of the shot too. So, a lot has been done and is in use regarding gunshots using digital systems, microphones etc. Pretty amazing stuff.
I'll have to do research on this too but if anyone can quickly read over my following idea(s) (especially the #2) to either shut it down with logic, or maybe say "theoretically, it's possible" then I'd really appreciate it:
Two microphones are available:
One is placed at the FIRST yellow figure, falling on the 90 degree line, closest to the red triangle (basically - in front of it)
Another is placed in the yellow figure falling on the 30 degree line
IF there was to be a gunshot in the red triangle - it'd first reach Microphone #1
and THEN reach microphone #2 (at a lower volume)
this is obvious, and I think that if you knew the positions of the two microphones (and synchronized them) then angle/distance could be achieved
but now..
What if I were to use the two (or three) microphones to measure how many shots were heard?
^This is the question that I *really* want answered (Since the previous one's answer seems easily doable)
Let's say there are
two guns - THIS TIME IN THE GREEN TRIANGLE .
One a bit more to the left, and one more on the right. (not too much, they need to fall within the green triangle)
They shoot "south", PARALLEL to the first 90 degree line.
if they were to shoot at the *same exact time* - then (please tell me if i'm wrong) this is what would happen:
Microphone 1 POV:
First - the shot from the GUN TO THE LEFT is heard (since that's closest to it)
THEN - the shot from the GUN TO THE RIGHT is heard (since that's the one further away)
Microphone 2 POV:
First - the shot from the GUN TO THE RIGHT is heard (since that's closest to it)
THEN - the shot from the GUN TO THE LEFT is heard (since that's the one further away)
so we have two audios that actually represent two different events.
would I be wrong to assume that, if "overlapped" or COMPARED in some sort of way
then they could give some information on how many shots there were?
second question related to this:
would it then be possible to tell if a shot was from the LEFT ONE or RIGHT ONE
by "overlapping"/comparing the two audios?
even though the overall amplitude of mic #2 is lower, the loudest shot for it would be the RIGHT one
while the loudest shot for the mic #1 would be the LEFT ONE
so how would one differentiate?
I know this may be a nonsensical statement but: would studying the phase(s) be of help here?
I apologize for these posts being long - hard to follow - and almost outlandish, I really am trying my best to explain my ideas and this is something that I definitely want to be able to pull off, it is of high importance to me
only 1 day has passed and everything which I've read here from everyone has made me learn so many things that it's mindblowing