README
======

Models in this directory were trained with the 

"NIST CD-ROM Version of the Texas Instruments-developed Studio Quality
Speaker-Independent Connected-Digit Corpus".


Contents of this directory:

am.props
--------

Properties file containing the parameters used to train the acoustic model.

wd_dependent_phone.500.mdef
---------------------------
Model definition file (ascii) which lists senone-ids and
HMM compositions

wd_dependent_phone.cd_continuous_4gau/
--------------------------------------

Directory containing one set of context-dependent acoustic models
(means, variances, transition_matrices, mixture_weights),
and the corresponding subvector-quantized models file (ascii)
with 200 codewords.

wd_dependent_phone.cd_continuous_8gau/
--------------------------------------

Directory containing one set of context-dependent acoustic models
(means, variances, transition_matrices, mixture_weights),
and the corresponding subvector-quantized models file (ascii)
with 200 codewords.

dictionary
----------
The training dictionary

fillerdict
-----------
The training filler words dictionary

unigram.lm
----------
ARPA format (ascii) unigram language model for the task.


Acknowledgments
===============

This acoustic model is trained from the TIDIGITS corpus, which is the "NIST CD-ROM Version of the Texas Instruments-developed Studio Quality Speaker-Independent Connected-Digit Corpus". 
