Extracting trigger data from real time environmental audio

Mar 3, 2008 at 4:07pm

Extracting trigger data from real time environmental audio

Time to ask for some help!

As part of a patch I’m working on, I want to use broad sound events
from the real world (cars/trains passing. people talking etc) to
create triggers to fire off different audio events and controllers of
those events.

I can look at overall sound levels easily enough using things like
avg~ ; but I wanted to see if I could get a little more subtle with
the analysis, and perhaps look at events happening in specific
frequency bands, both in terms of things happening at particular
blocks of frequencies, and also perhap transients.

In the past I’ve used a bank of resonant filters to do the activity
bit, but it’s crude, and I’m thinking that using fft~ and/or jitter
would be a better way to go. I’ve been playing with one of the eg
patches, where an fft~/cartopol combination is used to fill a jitter
matrix / jit.window. (If you know the eg patch, I’m using the
instantaneous values part of the patch).
That gives me a stream of lists giving the levels (0-255) of each band
of the fft analysis, and I’ve been using some of the LObjects to look
at activity in the lists, but I’m beginning to wonder if perhaps the
approach is wrong, and this could be done more easily in the fft
realm. However, I don’t really know where to start with that. I’m
also not sure if there isn’t a better way to so this in jitter than
the things I’m currently trying.

If anyone has any suggestions as to how to go about doing any of this,
where to start, what to think about, what objects I should use, or
anything else, I’d be really grateful. (If I haven’t given enough
detail about what I want to do, let me know)

thanks

David

#36160
Mar 8, 2008 at 3:37pm

look at tristan jehan’s [analyzer~]

M

On Mar 3, 2008, at 17:07, David Stevens wrote:

> Time to ask for some help!
>
> As part of a patch I’m working on, I want to use broad sound events
> from the real world (cars/trains passing. people talking etc) to
> create triggers to fire off different audio events and controllers
> of those events.
>
> I can look at overall sound levels easily enough using things like
> avg~ ; but I wanted to see if I could get a little more subtle with
> the analysis, and perhaps look at events happening in specific
> frequency bands, both in terms of things happening at particular
> blocks of frequencies, and also perhap transients.
>
> In the past I’ve used a bank of resonant filters to do the activity
> bit, but it’s crude, and I’m thinking that using fft~ and/or jitter
> would be a better way to go. I’ve been playing with one of the eg
> patches, where an fft~/cartopol combination is used to fill a jitter
> matrix / jit.window. (If you know the eg patch, I’m using the
> instantaneous values part of the patch).
> That gives me a stream of lists giving the levels (0-255) of each
> band of the fft analysis, and I’ve been using some of the LObjects
> to look at activity in the lists, but I’m beginning to wonder if
> perhaps the approach is wrong, and this could be done more easily in
> the fft realm. However, I don’t really know where to start with
> that. I’m also not sure if there isn’t a better way to so this in
> jitter than the things I’m currently trying.
>
> If anyone has any suggestions as to how to go about doing any of
> this, where to start, what to think about, what objects I should
> use, or anything else, I’d be really grateful. (If I haven’t given
> enough detail about what I want to do, let me know)
>
> thanks
>
> David

#123883

You must be logged in to reply to this topic.