Is it possible to use a sample recording of a voice, then using the negate (180 phase) of ambient or white noise as a "brush" of sorts to create the negative of the voice as a cutout of the ambient sound and actually create a pattern of silence that the human ear would perceive as a "voice"?
$\begingroup$
$\endgroup$
Add a comment
|
3 Answers
$\begingroup$
$\endgroup$
Not really.
Voice and out of phase voice sound exactly the same, so that wouldn't make a difference. It also wouldn't make a difference when overlaying with ambient sound. Ambient sounds and voice are typically uncorrelated, and hence phase makes no difference.
$\begingroup$
$\endgroup$
3
IMHO, this is an excellent question. I haven't heard of the experiment being done, though I think it could be a bit simpler than described in the question.
If one's brain expects to hear a certain noise spectrum and certain parts are missing, it seems likely that the missing parts will be noticed. And, if a certain spectrum is heard and then abruptly stopped, then the negative of that spectrum may well be perceived for some short time interval afterward, kind of like the negative afterimage that is perceived after glancing at a bright light.
The following experiment might be a good way to test your idea:
- Get voice prints (time-dependent spectra) for a short spoken phrase.
- Write a program to generate and send to a speaker a time-dependent broad-spectrum audio signal that is equivalent to noise with the voice print removed.
- Listen to the audio at different loudnesses.
If it works, you might find that it works better at high loudness. Or, you might find that it works better if you interleave full-spectrum noise with the voiceprint-subtracted noise, at intervals of a few tens or hundreds of milliseconds.
If it works in the scenario(s) described above, you can be pretty sure it would work if you can figure out how to modulate ambient sound the way you've suggested. But, frankly, doing that modulation could be much more difficult than it might appear at first glance.
-
$\begingroup$ Yes, you are right about the experiment. Maybe I shouldn't refer to experiments I have not cited. I'll do my best to find the corresponding paper. Otherwise, the process is pretty much what you describe, with the exception that you suggest the use of time-dependent spectra, while it would be easier (and maybe a better start) to use "static" spectra. Thanks for the answer though! A paper stating that the brain somehow uses the spectra of both ears is Pitch of Noise Signal - Evidence for a "Central Spectrum" by Frans A. Bilsen. $\endgroup$– ZaellixAFeb 12, 2020 at 10:50
-
1$\begingroup$ I suspect that changing spectra, a la voice prints, will be more perceptible than static spectra. Our sensory systems have a tendency to ignore unchanging stimuli. $\endgroup$ Feb 12, 2020 at 13:38
-
$\begingroup$ Yes, this may very well be the case. I thought that it would be quite easier to test with some static spectra but if it's run for some time, then using non-static spectra would be the best approach. $\endgroup$– ZaellixAFeb 12, 2020 at 14:20
$\begingroup$
$\endgroup$
Hilmar is correct here. Although you have to keep in mind that our hearing system has the tendency (or ability if you like) to compensate for spectrum irregularities. This is something that all of us experience every day when we move into a room that has "different acoustics" than outdoors (or the room we were before).
There has been some experimentation on this phenomenon. If I recall correctly, a sound with noisy spectrum (or it could be some kind of noise - pink, white, etc. -) was reproduced, then part of its spectrum was removed (or attenuated) and reproduced again for a while. After some time (some seconds I believe it was) the missing part of the spectrum was restored and then most people said that this part of the spectrum was too loud.
This shows that our hearing system makes some compensations and adaptations to the spectrum. Although this is not exactly what you asked for, you could possibly try to replicate the experiment with some noise and a voiced spectrum (subtract it from the noise and after a while restore it and check if it sounds like the voice you used).
EDIT
A paper proposing the existence of a "Central Spectrum" (as stated by the author) that results by some kind of averaging between the twoPitch of Noise Signal - Evidence for a "Central Spectrum" by Frans A. Bilsen. This somehow shows that the brain performs various averaging and or "optimization" process in order to extract information in an "optimum" (in the evolutionary sense I guess) way. Part of it should be the spectrum adaptation process.
