Is anyone aware of techniques or research on this?

  • possibly a cat@lemmy.ml
    link
    fedilink
    arrow-up
    12
    ·
    1 year ago

    It depends on your resonant cavities (among other things) but there are probably IRs out there that you could use to approximate it. IR stands for Impulse Response.

    I know I’ve seen at least one paper on this but it was probably over a decade ago and behind a paywall. You might try a search on ScienceDirect, Google Scholar, etc. But then you have to configure a process to match what the paper describes, which might not be trivial - each EQ curve causes phase dis-alignment, many processes are non-linear, etc. Getting a “flat” recording to work with would be very challenging by itself.

    • intensely_human@lemm.ee
      link
      fedilink
      arrow-up
      3
      arrow-down
      1
      ·
      1 year ago

      Here you go.

      You record the vibration of all the little receiver hairs in your ear as you speak. You simultaneously record into a microphone at the same time.

      You train a neural network to transform the sound file from the recording into the set of vibration states it corresponds with in your ear hairs.

      Then you just run the resulting model on the sound file, and magnetically vibrate the ear hairs in the way the model predicts.

      Boom! 1st person speaking voice experience.