0
$\begingroup$

I'm a biologist and I was thinking of ways that impulse responses could be used to simulate how an animal sounds over various distances in different types of forests.

Ultimately, if a tool was constructed that would allow scientists to simulate animal sounds over various distances, this would be a pretty useful research tool. Specifically, I'm interested in simulation of the reverberation effects, hence why I'm interested in impulse responses.

Process-based modelling an impulse response is not trivial, however, as an empiracle approximation I was wondering:

  • If I record an impulse response at 50m from a source in a forest, if I were to convolve a source signal twice, would that result in a reasonable approximation of what the sound would be at 100m distance?

The idea is that if I can accurately measure impulse responses over short distances in different forest types, these could be used to simulate reverberation over a large distance, including sound passing through multiple forest types. Currently, I can't think of why this would be a bad approach... but I'm a biologist, so I'm hoping others can tell me if this seems like a flawed concept.

$\endgroup$

1 Answer 1

1
$\begingroup$

if I were to convolve a source signal twice, would that result in a reasonable approximation of what the sound would be at 100m distance?

Yes. For small number of cascades this is a reasonable assumption.

However, you can't cascade the 10 meter response 10 times to get a single 100 meter response. While the general shape and structure of the response from 20m to 30m is the same as that from 30m to 40m, the phase spectra and the fine structures of the reflections are uncorrelatated. If you cascade the same impulse response over and over again, you get spectral thinning and it's starting sound "metallic" and "hollow".

You can try to decorrelate short repsonses from the original or just to start with a parametric model that has tree density, tree reflectivity, leaf/brush absorption and obvioulsy distance as tweakable parameters.

$\endgroup$
1
  • 1
    $\begingroup$ This was helpful! Can you elaborate on what exactly you mean by decorrelate short responses from the original? (I'm also quite happy with a link to a page on the topic if this is too much to explain in a comment) $\endgroup$
    – RTbecard
    Sep 30, 2020 at 16:09

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge that you have read and understand our privacy policy and code of conduct.

Not the answer you're looking for? Browse other questions tagged or ask your own question.