Audacity Forum

Audacity 2.0.5
Ubuntu 12.04

Hi all,
I'm new to all of this, leaning as I go, new to ubuntu, new to audacity.

The project I'm interested in developing is one where speech impared people, for example people who can only make grunts or sounds or poorly formed words, can have their grunts etc paired to a library of clearly spoken words. The person with the disability would have to spend time paring their grunts etc to the word they want to say. The clear word is then expressed through a speech synthesiser so that others can understand what they really intended to say.

My question is: can Audacity be used as the basis of this project?

Thanks in advance.

Ralph.

Audacity doesn't do anything in real time except Record, Play and Timer. So the best we could do is record the effort and try to pair it up with the appropriate match and then play the match -- minutes later.

I'm not sure the analysis software is up to this, either. One of our standing lines is Audacity doesn't speak English. It can't look at a collection of sounds and "know" that it's the Gettysburg Address. Word Recognition is actually a long-standing feature request. So there's no base on which to build the tools.

You'd be far better off with a word recognition engine linked to a text speaker.

And no, the Audacity developers don't relax all day. They have all they can do to correct errors and get the next version of the program out keeping in mind it has to be done in multiple languages and three major operating systems. Audacity is not an English Windows program.

If you know programmers and developers that can help out with the Audacity project, we'll be happy to talk to them.


Koz

An interesting project. Alas, it's beyond Audacity's capabilities.
At least, it would be to cumbersome for the user.
Let's take the simple case of sounds --> words (as text output). As Koz already said, the text had first to be recorded before the actual analysis. That's a thing one could live with.
However, the learning phase is nearly impossible to manage. All groans/sounds had to be recorded too and each one labelled.
This task gets very fast boring.
Ok, there may be the possibility to produce a Audio track with the desired words as a template (very easy with a text-to-speech program like "Balabolka".
The user can then record his interpretation inbetween the words. Very similar to a vocabulary learning tape.

The tricky part for the programmer is to identify those "Groans" and to assign them. The question is how many different sounds such a person can produce and how much they vary in pitch, loudness, timing, emotion etc.

It looks to me like a perfect task for an university that works with such prediction models.
A smart phone application would offer the greatest benefit imo.

"Praat" may be a better starting point for the proposed software: http://www.fon.hum.uva.nl/praat/