USE Impulse response to get Original sound

Question

I am a student majoring in acoustics. My English is not very good, so I apologize in advance. In my project, I am dealing with impulse response in room.

Find the impulse response function between the source and receiver in a given space using a microphone, speaker, and gunpowder. In theory, the impulse response between a source and receiver under the same conditions should have the same value. Of course, we know that the frequency characteristics of the sound source used by the source are not the same as a perfect impulse, so there will be errors.

Anyway, using the impulse response function I already have, I want to play another sound source at the source and deconvolve the data measured by the microphone at the same location with the impulse response function to get the original sound of the source. I want to get the same value in the time-domain, but I know that this is a difficult problem. So I want to have data similar to the frequency component of the original sound, and if possible, I also want to know the dB of the original sound.

In short, I want to remove the spatial characteristics of the impulse response function and find the "original sound".

There are several problems.

First is the selection of the sound source to use as a source to get the impulse response function. I have an omnidirectional speaker and its functions include white noise, impulse (implemented by the speaker), and sine sweep, but to get the impulse response function, I need to have data about the "original sound" coming from the speaker. I don't have it, so I am assuming that it is the "original sound" by placing a microphone nearby and processing the data, but the error is quite large, so it is a problem. If you know a good way, I would appreciate your help.

Secondly I need to deconvolve the data measured by the microphone with the obtained impulse response (using the iFFT method) and the data used as the source should be similar, but there is a lot of error. I would appreciate some theoretical help in this area.

"The data convolved with the impulse response function and the data actually measured at the location are somewhat similar.

However, I think that if the measured data is deconvolved to remove spatial characteristics, it should be similar to the "original sound", but it is not. I would appreciate if you could explain this in detail.

Thank you for reading this long question, and if you have any references, please let me know!!!

Translated with DeepL.com (free version)

Answer 1

using a microphone, speaker, and gunpowder.

Gunpowder sounds like a terrible idea. Both spatial and temporal properties of a gun shot are poorly defined and not particularly reproducible. Also the peak of the explosion is pretty much guaranteed to have a sound pressure level where the air becomes significantly non-linear.

we know that the frequency characteristics of the sound source used by the source

True, but this can be reasonably well corrected. The main problem here are the spatial characteristics (or radiation pattern) of the source. These are an integral part of the impulse response. There is no such thing as an "impulse response between point A and B". The impulse response you would get with a human mouth is different from a violin is different from a loudspeaker. There is no single "true" impulse response: they are all correct and it really depends on what you are planning to do with the impulse response.

This is true for the receiver as well: a human listener receives sound different from a microphone, but generally the differences are smaller here and good omnidirectional microphones are readily available.

In short, I want to remove the spatial characteristics of the impulse response function and find the "original sound".

That's impossible. The room transfer functions has a lot of zeros and "loses" information. It's not invertible. You can do partial corrections: like spectral equalization or dereverberation but this is complicated and often induces significant artifacts.

I would appreciate some theoretical help in this area.

Measuring room impulse responses is very difficult. Take a look at this answer for a discussion of different measurement methods: https://dsp.stackexchange.com/a/93637/3997 The most important thing here is that you need a quality metric to determine whether a measurement is good or not: specifically you need a good estimate of SNR and THD over frequency to determine over which frequency range the measurement is "usable".