I've noted in several places in the Sphinx docs from the CMU website that it's stressed that the AGC parameter of the decoder must match that of the acoustic model.
The Sphinx2 distribution contains a 6000 tied-states acoustic model in sphinx2-0.4/model/hmm/6k/. I found a description of this model at http://www-2.cs.cmu.edu/~rsingh/sphinxman/fr5.html, but this says nothing about whether this acoustic model was trained with AGC active or not. Can anyone tell me?
cheers,
jerry
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Anonymous
-
2003-03-05
The word from CMU is that it was trained with AGC set to 'none'.
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
I've noted in several places in the Sphinx docs from the CMU website that it's stressed that the AGC parameter of the decoder must match that of the acoustic model.
The Sphinx2 distribution contains a 6000 tied-states acoustic model in sphinx2-0.4/model/hmm/6k/. I found a description of this model at http://www-2.cs.cmu.edu/~rsingh/sphinxman/fr5.html, but this says nothing about whether this acoustic model was trained with AGC active or not. Can anyone tell me?
cheers,
jerry
The word from CMU is that it was trained with AGC set to 'none'.