VEDICS Speech Assistant

[rao.nischal]

VEDICS (Voice Enabled Desktop Interaction and Control System) is an assistive software which lets the user to interact with the OS using voice commands. VEDICS makes use of Sphinx 4 for doing speech recognition and hence inherits the cool feature of user-independence. That’s right!!! No training is required to use VEDICS!!
VEDICS lets you do the following with your system:

• Perform common window operations like close, minimize, maximize etc.
• Invoke default applications like browsers, mail clients etc.
• Access any element on the desktop just by saying its name.

We'll be adding the feature of dictation soon.

For more details visit: http://vedics.sourceforge.net/

The beta version of VEDICS is now available for download at http://sourceforge.net/projects/vedics/

Please give us feedback about your experience using VEDICS. This would help us in improving it.

For demos of VEDICS visit the channel: http://www.youtube.com/user/bharatj7

Thanks.
VEDICS Team

[rflores2323]

this is great. I am looking for something like this to be able to control my home theater pc with xbmc (www.xbmc.org). Couple of questions.

I see that you are able to open applications however are you able to open any application of just the default applications in ubuntu? I want to be able to open xbmc application up

Are you able to control an application by putting a voice command to a key command (keyboard emulation)?

also are you able to make custom commands? For instance making a prefix to be able to make sure the pc only hears the command when you say the prefix. For instance you say "jabba open xbmc" and it opens the application. But if you say open xbmc it does not do anything unless you give a special command like "jabba listen" so that it just listens to all your commands until another command is given.

basically take a look at this program that is currently freeware but only works on win7 http://voxcommando.com/
this app is really doing a great job of integrating with xbmc.

I am sure that you would get many followers from the xbmc community as its a large community and your software can really be added to it with the new add on plugin system they are developing. take a look at the forums and go to the linux forums as there are tons and tons of users.

[rao.nischal]

currently VEDICS supports only some default applications (browser, mail, text editor, terminal and help). For rest of the applications, you can navigate in the gnome panel to get to your application. If your application is gtk based, then you can access everything within that application.

VEDICS supports only some key commands (enter, escape, home end etc). We plan to provide full keyboard support soon.

hmm.. custom commands is a nice idea. Again we will try to get that in the next release.

[Automat2]

Watching the youtube video, you can think that for people like me, who have a strong foreign English accent, the voice recognition by the software would be difficult. It happens several times on the video.

I remember when I used Dragon, had the feature of "learning your voice", so it facilitated the "understanding" of your voice by the software. Is that present on Vedics?

[rao.nischal]

Yes, it might be slightly difficult to do perfect speech recognition with a strong accent. VEDICS is expected to work well with US English accent (since VEDICS makes use of Sphinx 4). However, you may have noticed in the videos that the speaker's accent isn't what you'd call a US English accent. And still it works to some extent (may be with a lower performance). All this happens due to the limitation that VEDICS uses Sphinx as speech to text converter.

VEDICS currently makes use of Sphinx4 (it is still beta) as the speech-to-text converter. As far as I know no other speech-to-text converter for Linux provides as good a performance as Sphinx does. One of the objectives of Sphinx is to make speech recognition speaker independent. They have tried making it as independent as possible however it still needs a lot of improvement.

To make a perfect speech-to-text converter, very good acoustic models are needed.And to get reliable acoustic models lot of speech samples are needed. One way in which you can help building a reliable speech recognition system is by donating your speech. You can do it online at: http://www.voxforge.org/home/read

The more speech samples the better. So donate as much speech samples as possible so that you get performance in speech recognition.

Sphinx currently doesn't make use of "learning your voice" feature (because they are trying to make speech recognition speaker independent).

[dansan]

VEDICS (Voice Enabled Desktop Interaction and Control System) ...

We'll be adding the feature of dictation soon.

...
VEDICS Team

This is great news, Rao. Lack of speech recognition in linux, specifically dictation, has been the only deal breaker for me to switch my translation business entirely to Ubuntu.

I would gladly donate time to help with this project, for example, in providing speech samples (American English) for the language model or documentation.

Best regards and thanks,
Daniel

[dansan]

...
To make a perfect speech-to-text converter, very good acoustic models are needed.And to get reliable acoustic models lot of speech samples are needed. One way in which you can help building a reliable speech recognition system is by donating your speech. You can do it online at: http://www.voxforge.org/home/read...

I can't get the above web page to display the recording applet. I'm running Lucid Lynx with FF with all variations of java installed. FF Advanced Preferences has Enable Java Scripts ticked.

Any clues about what's wrong?

[rao.nischal]

you need to have java plugin for firefox to run the applet.

[dansan]

you need to have java plugin for firefox to run the applet.

Thanks for this tip, Rao. After following it, I get the applet to open but still no joy because when I click a Record button, the Java applet displays a message, saying I don't have permission to record.

java.security.AccessControlException: access denied
(javax.sound.sampled.AudioPermission record)

So far I have found some solutions calling for adding code to .java.policy, but I haven't found out how to do this, so I asked for help in the Multimedia & Video Forum.

Meanwhile, on my Windows 7 box I have none of these problems. So, I've posted some speech samples using it. I gather from discussion on that site that 15 minutes of samples is a useful contribution. Is that right?

[rao.nischal]

Thank you very much for submitting speech sample. Any amount of speech sample submitted is always better than nothing.