Dialog Managers

Flat
acoustic model gnome-voice-control
User: Robin
Date: 1/18/2008 1:16 pm
Views: 17934
Rating: 38

The results I got using Gnome-voice-control were worse than I expected after seeing the demonstration videos. I could only use about three commands successfully.

I was wondering whether this could be caused by the acoustic model? I probably used the sphinx2-hmm-6k acoustic model, because I installed gnome-voice-control under ubuntu directly from the repostories. 

Was there another acoustic model used for the screencast?  For instance one that was especially adapted to Raphael’s voice?

Will it be wise in my case to use the model from VoxForge?  (I have donated about an hour and a half of speech) If so, how do I replace it?

Thanks for any pointers!

Robin

--- (Edited on 1/18/2008 1:16 pm [GMT-0600] by Robin) ---

Re: acoustic model gnome-voice-control
User: nsh
Date: 1/18/2008 1:35 pm
Views: 1423
Rating: 37

> The results I got using Gnome-voice-control were worse than I expected after seeing the demonstration videos. I could only use about three commands successfully.

Hm, that's strange. Did you enable accessibility? Actually you need to enable it in order to move through interface. Without accessibility only a few commands will work.

If it will be broken still, please record a few commands and send them to me ([email protected])

> Will it be wise in my case to use the model from VoxForge?  (I have donated about an hour and a half of speech) If so, how do I replace it?

About voxforge model, it's a very important task really and should work better for you. But nobody created voxforge for sphinx model up to now. David promised me to run it on CMU cluster once setup we'll be prepared (27 hours should process about a week on usual machine) but someone should sit down and prepare configuration for sphinxtrain. Also if we'll have voxforge model we can do real tests with it (David can measure it on standard wsj and hub test sets I suppose).

 

--- (Edited on 1/18/2008 1:35 pm [GMT-0600] by nsh) ---

Re: acoustic model gnome-voice-control
User: Robin
Date: 1/19/2008 5:50 am
Views: 343
Rating: 46

I went to:

preferences>Universal Access>Assistive Technology Preferences 

There I checked the checkbox for:

Enable assistive technologies

Is this what you mean? If not can you give me the names of the menu items.

It still only recognizes

- minimize window

- next window

- close window

With "previous window" it starts evolution.

For the rest nothing (e.g. "file", " edit" etc) seems to work.

Tell me if I missed something... if not I'll record some commands. Do you want menu entries such as "file", "edit" etc?

--- (Edited on 1/19/2008 5:50 am [GMT-0600] by Robin) ---

Re: acoustic model gnome-voice-control
User: nsh
Date: 1/19/2008 6:10 am
Views: 293
Rating: 37

> Enable assistive technologies

Right, but you have to relogin after that. Also please check that you are using C locale, native local won't work. i.e. gedit should have File menu.

> With "previous window" it starts evolution.

Right, as far as I remember there was no previous command and no OOV filter

> For the rest nothing (e.g. "file", " edit" etc) seems to work.

If "next window" command works fine, the most probable reason is missing ally. You can try accerciser to check if allyis present and can explore UI.

--- (Edited on 1/19/2008 6:10 am [GMT-0600] by nsh) ---

Re: acoustic model gnome-voice-control
User: Robin
Date: 1/19/2008 12:17 pm
Views: 333
Rating: 37

I even rebooted, because I thought I remembered having read something like that.

I'll check out the rest as soon as I'm home again.

> i.e. gedit should have File menu

You mean simply a standard menu File etc? I think I have that too.

So I probably miss ally. Never heard of accerciser though... Where  do I find that?

Thanks,

Robin 

--- (Edited on 1/19/2008 12:17 pm [GMT-0600] by Robin) ---

Re: acoustic model gnome-voice-control
User: Robin
Date: 1/21/2008 3:35 am
Views: 325
Rating: 35

Found and installed accerciser. It froze when I started it. I'll try again next weekend. I'll let you know my findings.

 

--- (Edited on 1/21/2008 3:36 am [GMT-0600] by Robin) ---

Re: acoustic model gnome-voice-control
User: nsh
Date: 1/21/2008 11:03 am
Views: 353
Rating: 36
Please do. It should be rather straightforward really.

--- (Edited on 1/21/2008 11:03 am [GMT-0600] by nsh) ---

Re: acoustic model gnome-voice-control
User: Robin
Date: 1/27/2008 11:00 am
Views: 304
Rating: 34

> You can try accerciser to check if allyis present and can explore UI.

Accerciser now works, but how do I use it to check if ally is present?

--- (Edited on 1/27/2008 11:00 am [GMT-0600] by Robin) ---

Re: acoustic model gnome-voice-control
User: nsh
Date: 1/27/2008 11:12 am
Views: 317
Rating: 38

If you can explore gedit UI with it, everything is fine.

 

--- (Edited on 1/27/2008 11:12 am [GMT-0600] by nsh) ---

Re: acoustic model gnome-voice-control
User: Robin
Date: 2/2/2008 5:53 am
Views: 307
Rating: 36
Looks okay I think ... I'll send you 2 screenshots. One of gedit and one of accerciser exploring gedit. Let me know if it looks like I checked the things needed.

--- (Edited on 2/2/2008 5:53 am [GMT-0600] by Robin) ---

PreviousNext