Re: Request for Assistance Finding FLOSS Speech Recognition Software

[meg ford <meg387 gmail com>]

Hi,

On Wed, Jun 27, 2012 at 3:25 AM, Peter Grasch <grasch simon-listens org> wrote:

Hi,

Am Dienstag, 26. Juni 2012, 17:07:04 schrieben Sie:

> "That sounds good. This is the best thing I 've heard about, because of the
> flexibility. I don't really understand what he meant. It wasn't clear to me
> if he meant that you would have to add programming code of if he just meant
> you have to train everything.

No, you don't need to write any code. But you need to define your own
vocabulary, grammar, commands and trainings texts. There are pre-built
packages for this ("scenarios") already available but for such custom
installations we recommend to build one according to the users needs and
pronunciation (often there are certain words that are easier to pronounce
consistently than others).

> I 'm not even going to try to get it to
> understand words at this point -- I just want to do the alphabet.

Perfect. That's certainly doable. There a pre-built scenarios online for the
virtual keyboard (roughly) following the phonetic alphabet.

I'm not sure though if it's good to start with so many (relatively) words
right away (I think it's around 35 words or so). A Dasher setup would lead to
good results much quicker (only around 7 word).

 

Can you explain this? Are you suggesting that he use Dasher for word prediction?

But it's mainly a question of how much training is possible.

> Anyway,
> I'll just download it and see how it works. I might try both the Spanish
> and the English to see if one works better than the other."

I think both languages should work equally well. There is no bias towards any
language because you'll build your own acoustic model. It doesn't matter to
Simon if this is English, Spanish, French or Chinese. You can even mix
languages if you want (take care with the phoneme sets in this case, though).

I'd suggest you pick the language you are more comfortable with and - again -
can pronounce more consistently (remember, you're going to call out commands
spontaniously during "normal" usage).

This was the main problem with other programs. Joe has around three distinct ways of pronouncing the same word, depending on his muscle tone.

> I (this is Meg speaking again) am currently working on GNOME through Google
> Summer of Code. I will see if I can help Joe get started when I am done
> with GSoC. Thanks very much for the information, Peter. I will contact you
> when he gets started.

Setting up Simon is not that hard considering there is a lot of documentation
available.
Still, setting up Simon *well* for people with dysarthric speech can be a lot
harder as there is a lot of potential for improvement from a stock
installation. For example, a custom phonetic transcription can improve
recognition rates tremendously (we went from around 45 % recognition rate to >
95 % simply by using custom transcriptions for one of our test users suffering
from cerebral palsy).

That's why we do also offer consulting through the Simon Listens e.V. where we
basically build you a custom speech model for use with Simon according to your
needs for a modest fee to cover our expenses (non profit organization).

This is potentially something he is interested in, yes. He may have to apply for funding for this, however. Would you need to travel to the United States to do this? If you have standard information available about this, I would be grateful if you would email it to me directly.


Thanks!
Meg Ford

Let me know if this is something you'd be interested in.

Best regards,
Peter