h a l f b a k e r yNo, not that kind of baked.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
Current speech synthesis creates a voice that sounds robotic and lifeless. I was reading somewhere that some computer scientist said that we'll probably have to model the hole nose, mouth, throat, and lungs in order to get realistic speech synthesis out of a computer.
I propose something that is
more robotic. This device creates a vibration mechanically, and complete with a computer controlled robotic larynx, cheeks, tongue, and cheeks, it shapes the sound coming out to make a more realistic speech.
Erasmus Darwin created an early speech synthesis device with a silk ribbon for the larynx, and lips made out of leather. It was able to say 'map', 'pam', 'mama' and 'papa'. Presumably with finer motor control, we could get such a contraption to say anything.
Erasmus' early device
http://www.stanford...epresentations1.pdf See page numbered 105, actually 9 of 30. [lawpoop, Jan 15 2005]
Mechanical speech device, built for the Exploratorium science museum
http://www.explorat...s/vocal_vowels.html Long ago I watched a documentary about the making of this exhibit. [robinism, Jan 15 2005]
Alexander Graham Bell
http://www.pbs.org/...eevents/mabell.html Right. [contracts, Jan 15 2005]
Much Better Link
http://www.ling.su....hartmut/kemplne.htm The History of Speech Synthesis [contracts, Jan 15 2005]
Hideyuki Sawada
http://www.eng.kaga...sawada/index_e.html This guy is on it like white on rice. [tiromancer, Jan 15 2005]
Visual Voice
http://www.visual-voice.com/ Interesting byproduct of looking for that guy [tiromancer, Jan 15 2005]
uncanny valley, population 0.999999
http://en.wikipedia...wiki/Uncanny_Valley The concept [bristolz] refers to. [jutta, Jan 17 2005]
mechanical birdsong synthesis
https://www.youtube...watch?v=tPKFT_t2rL0 [notexactly, Dec 21 2015]
Preheated - in William Gibson's " Neuromancer "
http://www.technove...content.asp?Bnum=72 [normzone, Dec 21 2015]
Baked.
https://www.youtube...watch?v=HmSYnOvEueo I would have added this link yesterday but didn't feel like searching for it. [notexactly, Dec 23 2015]
[link]
|
|
very interesting, but not useful. Perfect for the halfbakery [+] |
|
|
Not only was it baked by whomever "Erasmus Darwin" was, but Alexander Graham Bell also baked this in the eighteen hundreds. [link] exerpt: //Working out of their home, the industrious pair created an apparatus consisting of a facsimile mouth, throat, nose, maneuverable tongue, and bellow lungs.// |
|
|
I'm pretty sure this doesn't constitute a new idea, although it would be a good project for a thesis. |
|
|
This idea is different though, because it is computer controlled, and because it is used as part of a robot. |
|
|
Those early devices could only say preprogrammed things. You had to set the device up to say "mama" or "map". Erasmus' early dvice had a vocabulary of only 4 words. |
|
|
This device could say any electronic text. So this would be great for people who need speech synthesis, like Stephen Hawking, or for computer chat or something. For reading text, it's not new, but it's an improvement. |
|
|
Speakers already do a near perfect job of immitating humans. The problem lies in the computation of phonetic sounds from a text source. What you propose is nothing more than a very elaborate speaker, which still suffers from the aforementioned problem. |
|
|
[Aq-Bi] A mechanical device remove a lot of the computation. If you look at the duck calls imitating human vowels in the second link, you will see there is not a lot fo computation going on there. |
|
|
If you have a plastic box that is used as a larynx, you just saved yourself all the computation that goes into modelling a larynx. |
|
|
If you have some kind of rubber or plastic tongue, you don't have to calculate the tongue itself, you just need to calculate where to move it. |
|
|
So mechanical organs remove a lot of the aforementioned problem. |
|
|
BTW, am I the only guy who thinks some of the older speech synthesizers sounded better than the newer ones? The older ones had a 'blander' inflection, but one which was usually right. The newer ones try to impart more subtle inflections, but they're more apt to be noticeably 'off'. |
|
|
"Hello. I am Doctor Sbaitso. I am here to help you." |
|
|
Sbaitso was okay, but not all that great. Macintalk was better. |
|
|
I agree with you, [supercat]. Perhaps we've entered the uncanny valley? |
|
|
How about a new text where the letters are shaped like
those plastic vocal tracts in the second link? Or better,
where the serrifs on the letters are the parts of the vocal
tract most responsible for the sound and they are
exagerated in size. |
|
|
I have done tests with a Voltage Controlled Oscillator driven by a phototransistor aimed at the mouth that was lit by LEDs. There seems to be lip/teeth information there for vowel and consonants. I'm saying that this could be additional information for the speech synthesizer from a speaking person origin only. |
|
|
It would be great to have a version of this completely
manually controlled -- so maybe build a rubber mouth and
vocal tract and have one hand control the lips and teeth
and the other the tongue? And then put the lungs
between your knees so you could control volume, or
under the arm like a bagpiper -- thats it, this would be an
acoustic theramin bagpipe vocalizer! People could get
really good
at singing with whatever kind of a voice they wanted to
create. I have created a really minimal version of this by
cupping a circuit board lie detector in two hands and
learning to "mouth" the phrase, "ma ma!" |
|
| |