1
$\begingroup$

I would like to know how much of the sound of my voice getting through various types of walls can be reconstructed with easily obtainable software and algorithms.

My plan is to place a microphone on the other side of the walls I'm interested in and record the sound getting through. I would then apply reconstruction and try to understand my own speech. Since I am the one talking and I know what I was saying, this would be best case.

Are there ready algorithms, as standalone software, libraries or easy to implement in common software packages like Python, Matlab, Octave, ... (I'm talking about script-kiddie level of knowledge) that I can try to reconstruct offline the speech from the distorted recording?

$\endgroup$
1
$\begingroup$

This is a difficult problem. The wall itself is simply a linear time invariant filter. Typically it's just a low pass filter and can theoretically be compensated with a matching inverse filter.

However in practice the attenuation at most frequencies is quite severe. Unless the recording room is extremely quiet and the recording extremely low noise, the remaining speech signal will be below the noise floor and hence impossible to recover.

$\endgroup$
  • $\begingroup$ My experience was that I was able to guess some words when people talk in the next room, even if the sound is very distorted. I thought the brain does something like in image processing is called "super resolution", completing the missing parts using its previous knowledge. I thought there was something similar for audio. If that's not possible, even better. $\endgroup$ – FarO Apr 26 '16 at 21:28

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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