Acoustic Model Discussions

Nested
Error in "Test Acoustic Model Using Julius"
User: kay
Date: 3/1/2008 6:36 am
Views: 5963
Rating: 17

When testing the acoustic model I ran the instructions: 

julian -input rawfile -filelist wavlst -smpFreq 48000  -C julian.jconf  > juliusOutput  (copied from "step2-creating test prompts" )

but did not get the correct results.The error infomations  were:

--------------------------------

###### check configurations
###### initialize input device
###### build up system
Reading in HMM definition...(ascii)...limit check passed
   defined HMMs:    50
  logical names:   506 in HMMList
    base phones:    44 used in logical
done
Making pseudo bi/mono-phone for IW-triphone...369 added as logical...done
Reading in dictionary...
line 18: triphone "*-z+ih" or biphone "z+ih" not found
line 18: triphone "z-ih+r" not found
line 18: triphone "ih-r+ow" not found
> 6     [ZERO]  z ih r ow
error in reading sample.dict: 1 words failed out of 18 words
ERROR: failed to read dictionary, terminated
Terminated

------------------------------ 

i compare my tiedlist with the one given in "Step 10 - Making Tied-State Triphones",and they are completely the same,and files sample.grammar and sample.dict are also copied from "step2-creating test prompts".

That is why? 

--- (Edited on 3/1/2008 6:36 am [GMT-0600] by kay) ---

Re: Error in "Test Acoustic Model Using Julius"
User: kmaclean
Date: 3/1/2008 10:52 am
Views: 308
Rating: 25

Hi kay,

The pronunciation dictionary used in the Tutorial and How-to is based on the ISIP Switchboard corpus (contains around 27,500 words).  Whereas the  QuickStart and nightly AM builds is based on version 0.6 of the CMU Pronunciation Dictionary (contains around 130,000 words).   Unfortunately, the Switchboard and CMU pronunciation dictionaries use slightly different phoneme syntax.  This is enough to make them incompatible from a Grammar and Acoustic Model testing perspective (see ticket #52).

Did you select the correct sample.voca file in Step 2 - Create Test Prompts?  Your voca file pronunications need to match those  in the pronunication dictionary you used.   In this case, does the pronunciation dictionary you used to train your acoustic models use the same phonemes that you specified in your voca file?: 

[ZERO]  z ih r ow

Ken 

--- (Edited on 3/1/2008 11:52 am [GMT-0500] by kmaclean) ---

Re: Error in "Test Acoustic Model Using Julius"
User: kay
Date: 3/2/2008 1:09 am
Views: 2570
Rating: 23

Thanks a lot.I check my sample.voca and find the inconsistence in voxforge_lexicon which is a modified version of the Pronunciation Dictionnary included with ISIP Switchboard corpus and sample.voca.And I use the lexicon to build the acoustic model and  it can run now.

ZERO is defined as:

z iy r ow  (in voxforge_lexicon)

z ih r ow  (in my original sample.voca) 

--- (Edited on 3/2/2008 1:09 am [GMT-0600] by kay) ---

PreviousNext