7
$\begingroup$

We know from the International Pronunciation Alphabet (IPA) and the research that gave us the IPA, what sounds typically used in human speech (certainly not all sounds that can be made). This is all well and good but I want to do this for aliens with arbitrary physiology.

I don't want approximations or guesses. I'm looking for a software suite that might be adapted to my task. If the software works best in the hands of a PhD student or post-doc, that's probably the level of accuracy I'm looking for. I don't care about what sounds an alien will make, only the ones their physiology will allow them to make. If there are lower accuracy, but more approachable packages, I'd be interested in them too.

Computational complexity isn't a big problem as AWS/GCE is cheap and I have time.

(I also am not afraid of ruining monster movies for myself when some huge monster makes a noise pitched far above what their physiology would permit.)

$\endgroup$

This question asks for hard science. All answers to this question should be backed up by equations, empirical evidence, scientific papers, other citations, etc. Answers that do not satisfy this requirement might be removed. See the tag description for more information.

  • 2
    $\begingroup$ The IPA alphabet does not transcribe "all sounds human physiology is capable of making", but only those sounds which (1) are used in normal speech and (2) are considered sufficiently distinct. For example, IPA cannot represent music (at all), cannot represent whistling (at all), and of the various noises made by humans it represents only the few which are actually used as phonemes in some language somewhere. Moreover, the definitions of IPA symbols are not done in terms of physical sound, but in terms of human articulation, e.g., /ɔ/ is defined "open-mid back rounded oral vowel". $\endgroup$ – AlexP Dec 20 '17 at 1:15
  • 2
    $\begingroup$ If you're looking for that kind of software, you will probably have to pick what physiologies your alien might have. I'm thinking of how specialized the Vocaloid synthesizer is, and how complicated it would be to extend that to an arbitrary physiology. $\endgroup$ – Cort Ammon Dec 20 '17 at 2:14
  • 2
    $\begingroup$ I would be very surprised if any such effort had ever been made. The human phonatory apparatus is quite complex, and it is very far from being the only kind of phonatory apparatus used by terrestrial animals. For example, a bird's syrinx is very different, and a grasshopper's stridulation is utterly unrelated. I can't see how those three examples (mammalian larynx, tongue, lips, teeth and nose, avian syrinx, and insect stridulation) could possibly be subsumed in one single set of parameters. $\endgroup$ – AlexP Dec 20 '17 at 2:17
  • $\begingroup$ @CortAmmon that's a good pointer, thank you. The closest I've been able to find this far is to hack out something in Matlab and/or ANSYS. I'm also aware that this could be a PhD and post-doc project in and of itself. $\endgroup$ – Green Dec 20 '17 at 2:17
  • $\begingroup$ @AlexP they probably can't and so I'd have to develop a different model for each type and physiology I came up with. The math is already getting very complicated. $\endgroup$ – Green Dec 20 '17 at 2:20
7
$\begingroup$

Probably doesn't exist

...but here's what it would need to be able to do.

Primary Goal: Finding the vocal range of a arbitrary phonatory apparatus is the primary use of this software. We want to be able to design an apparatus, hit "Go" then come back and find out the vocal range possible. As much of the tedious modeling and data handling work should be done in software, not by the operator.

Requirements

  • Build a library of biological materials such as muscle, fat, bone, tendon, ligament, organs, fascia, mucous membranes and so on. Be able to assign these materials to various parts of the model. Have sane defaults for all materials (based on Earth analogs) but also permit custom tweaking of those parameters.
  • Be able to infer the relationships between various body structures. For example, airways are usually surrounded by skin then rings of cartilage. Fascia connects all that together. Modeling all that by hand is tedious in the utmost. Software should be able to do that for us.
  • Reduce the absurd number of degrees of freedom in a soft tissue problem like this to something more manageable. (No need to use 1000 CPUs if you can get by with 100.)
  • Define atmospheric parameters.
  • Rapidly vibrating muscles are frequently a part of a phonatory apparatus. The software should be able to simulate the interactions between these vibrating muscles and the surrounding tissue.
  • Account for changing geometry of the airway due to muscle and skeleton position. Human singers sound different when they hold their head a slightly different way. The software should be able to detect and determine the difference.
  • Account for resonance with skeletal cavities. We should be able to test the effects of the nasal cavity of Parasaurolophus

Secondary

  • Find overlap between sounds possible with sounds that humans can make. For example, if we define an alien that can use some of the same sounds as humans, we'd like to know that quickly.
  • Employ neural networks to drive the muscles to make sounds and speech.

Papers of Interest

I'm sure there will be many more papers to be read.

$\endgroup$
  • 3
    $\begingroup$ This answer fails to meet the expectations of the hard-science tag. It also doesn't answer the question, just lays out suspected requirements of the desired software. Not sure why you answered your own question with a non-answer... $\endgroup$ – Frostfyre Dec 20 '17 at 14:08
  • 2
    $\begingroup$ @Frostfyre I answered my own question because I'm fairly sure that the software I'm looking for doesn't exist and I'd have to write it myself. In lieu of existing software, i wrote down the specifications. $\endgroup$ – Green Dec 20 '17 at 15:06

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.