0
$\begingroup$

Are there any sample Audio files, encoded in the simplest format known, that can mathematically explain or illustrate the fundamentals of (audio) signal processing that might also relate to known equations of physics pertaining to audio?

I am trying to teach a robot AI system how to sing.

Improve this question
$\endgroup$
2
  • $\begingroup$ This seems somewhat vague and difficult to address, but let me start off with a few things. The pattern helps test TV image production and alignment in an effort to calibrate for the reproducibility of an image given a coding format. The TV is not being given any information about the nature of light. It is merely a way to ensure the hardware and software is correctly decoding the analog or digital image data. In addition, we perceive sound and light in dramatically different ways, visible light being observed using receptors sensitive to the primary colors and total luminance (1/2) $\endgroup$
    – Keegs
    Dec 30, 2020 at 3:27
  • $\begingroup$ Sound is interpreted through resonance in a complex structure of your inner ear that decodes sound based on its frequency content at no fixed number of frequencies (unlike light’s three). In TV it makes sense for a set of 6 primary and secondary colors as well as white and black, but there is no equivalent in sound. We perceive sound as a roughly continuous spectrum. (2/2) $\endgroup$
    – Keegs
    Dec 30, 2020 at 3:32

1 Answer 1

Reset to default
1
$\begingroup$

If you want the AI to relate to how we perceive sound, maybe something like

  • a linear chirp from 20Hz-20kHz
  • white noise in the same bandwidth
  • a single sine wave at 1kHz (roughly our greatest sensitivity)
  • a harmonic series from 55hz doubling up to 14.08kHz (all A notes to illustrate octaves)
  • other harmonic series and/or progression to illustrate important musical intervals (major and minor chords, pentatonic scale)
  • sound level variation of one or more of the above between 0dB SPL and some appropriate limit to establish dynamics

In terms of storage, a set of vectors is probably the best place to start. You can make one long vector with your test pattern and load it into and out of a WAV file I suppose.

Improve this answer
$\endgroup$

Your Answer

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

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