« How can I export a sam speech model?
shadowvocabulary.xml »

Static model from Voxforge

It is getting better:
1. I set the paths to the Voxforge acoustic model (English) files using simon > Configure > Model Settings > Static model > hmmdefs / tiedlist / macros / stats.
2. I imported the dictionary file dict as HTK lexicon into the simon active vocabulary.
3. I started ksimond.
4. It was possible to use the Synchronize button. simon should now be able to recognize English words that are part of the Voxforge English speech model because simon is connected and activated. I have to find out which microphone / headset I can use.

Tags:

This entry was posted on Wednesday, February 3rd, 2010 at 5:23 pm and is filed under Ubuntu. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.

3 Responses to “Static model from Voxforge”

  1. producer says:
    February 3, 2010 at 6:06 pm

    I think that I forgot to use the dictation plugin (Commands > Manage Plug-Ins > Add > Dictation). Now, with the dictation plugin, I get recognition results. It recognizes only DESIGNER and DESIGN, but at least simon recognizes something. That is a start!

    It is really good to see that it is possible to use the Voxforge English acoustic model with simon.

  2. producer says:
    February 3, 2010 at 6:26 pm

    Interesting: The file /home/am3msi/.kde/share/apps/simon/model/model.dict consists only of 4 lines:

    0 [] sil
    1 [] sil
    2 [DESIGNER] d ix z ay n er
    2 [DESIGNS] d ix z ay n z

    And my 200 words German speech model contains exactly these two words (this demonstrates that english pronunciation is possible with Ralf's German dictionary when it comes to foreign words like designer).

    It seems that model.dict is the intersection of my previous German active vocabulary with 200 German words and the imported English dictionary from Voxforge.

    So something went wrong. I don’t know why the intersection happened. But I think that I should completely delete the following folders:

    /home/am3msi/.kde/share/apps/simon
    /home/am3msi/.kde/share/apps/simond

    After I have done that, I can import the English Voxforge acoustic model again. Or I could try to restore my 200 words German speech model. Or I could start from scratch with my German speech model. I don’t know what I will do next.

  3. Peter Grasch says:
    February 3, 2010 at 11:02 pm

    Hm let me know what happened if you find out.

    Thanks!