speech synthesized from hand movements

If we had a setup to control parameters of synthesized speech
in real time, and a glove input device, we could do this with
human input and model it.

Avery, have you found anything for tweakable speech synthesis?
I haven't yet--or the literature I have looked at doesn't say so
in so many words.

        Bruce

[Martin Taylor 950112 12:30]

Bruce Nevin, apparently 12 Jan 1995 10:32:01

I don't know whether Bruce's message got damaged or whether it was intended
as a private response to someone, but here's how it starts:

If we had a setup to control parameters of synthesized speech
in real time, and a glove input device, we could do this with
human input and model it.

What is "this" that we could do?

There has been a project at the University of Toronto to synthesize speech
from hand movements. I saw a brief demo two or three years ago, and
a couple of days ago there was a segment about it on a science program.
It can be done.

Avery, have you found anything for tweakable speech synthesis?
I haven't yet--or the literature I have looked at doesn't say so
in so many words.

What do you mean by "tweakable?" All speech synthesis is tweakable, at
least insofar as all the parameters are variables that define the sound
emitted at any particular moment. There have been demonstrations of
changing the nature of the spoken voice by particular parameter changes--
one great demo by Melvyn Hunt (done quite a few years ago) has the
sentence "We were away a year ago" spoken by the Berlin Symphony as the
first bars of Beethoven's Fifth Symphony!

Martin

(Bruce Nevin Thu 930112 17:20:10 EST)

[Martin Taylor 950112 12:30]

What is "this" that we could do?
What do you mean by "tweakable?"

Martin, what I'd like to be able to do, for those parameters of speech
that people are observed to vary when speaking, is to put corresponding
parameters of synthesized speech under people's control by other means,
e.g. by hand movements. I am interested in phonemic contrast
(differences that make a difference between words) rather than the many
other parameters such as voice quality that are not phonemicically
contrastive. I would like to be able to disturb one or a combination of
such parameters, verify which disturbances people in fact do resist, and
track what people do to resist them. I would like to develop control
systems that model that. Hence, the need for the "tweaking" to be in
real time, and accomplishable by e.g. mouse movements or (if necessary)
something like a glove input device.

If there is equipment available, preferably at low cost, ideally
public-domain DOS software, that can support this kind of research, I
would like to know how to get it. Avery was going to look, and I have
done some looking. Perhaps you can direct us right to what we need.

Thanks,

        Bruce

[Martin Taylor 950113 10:50]

Bruce Nevin Thu 930112 17:20:10

[Martin Taylor 950112 12:30]

What is "this" that we could do?
What do you mean by "tweakable?"

I am interested in phonemic contrast
(differences that make a difference between words) rather than the many
other parameters such as voice quality that are not phonemicically
contrastive.

OK. I understand. That's harder than the voice-quality effects I was
thinking of.

If there is equipment available, preferably at low cost, ideally
public-domain DOS software, that can support this kind of research, I
would like to know how to get it.

I don't know, but maybe I could find out something about the U of Toronto
project on synthesis by Data Glove. It's a long shot, but a possible.
Also, I could ask my colleagues on the NATO Research Study Group on
automatic speech processing. They might know something. How serious are
you about it? Interested and curious, or prepared to put some effort into
making something work?

Martin

(Bruce Nevin Fri 930113 18:06:20 EST)

[Martin Taylor 950113 10:50]

How serious are
you about it? Interested and curious, or prepared to put some effort into
making something work?

Determined to put some effort into making something work, but prepared?
That suggests I have a way around persistent limitations of time and
resources, and I don't. Completing the phonology dissertation is a
(back-burner) priority, but I entertain the hope that PCT research
results could be worked into it; that would reduce conflict.

Avery, Bill, and I have had some off-line discussion of getting something
going. Initial effort would be very limited by design. One change: I do
now have a 386 desktop PC with mouse--my old 286 died abruptly, and I got
this used for my wife to work on a book.

Off until Monday.

     Bruce