Click here to register.

Step 8 - Realigning the Training Data

Background 

This operation is similar to the HLEd word-to-phone mapping operation performed in the Step 4, however in this case the HVite command can consider all pronunciations for each word (in the case where a word has more than one pronunciations), and then output the pronunciation that best matches the acoustic data.

Steps 

Execute the HVite command as follows:

$HVite -A -D -T 1 -l '*' -o SWT -b SENT-END -C config -H hmm7/macros -H hmm7/hmmdefs -i aligned.mlf -m -t 250.0 150.0 1000.0 -y lab -a -I words.mlf -S train.scp dict monophones1> HVite_log

This creates the aligned.mlfaligned.mlf file.

Review the output of the HVite command very carefully.  Catching errors here will save a lot of headache later on.   Because seemingly minor problems at this step sometimes show up as major errors at later steps, and they are very difficult to trace back to here.  Here is the log output from the above noted command: HVite_loghvite_log.  It is time well spent to review the log to make sure that HVite recognized all the words for each line in your prompts file.

Next run HERest 2 more times: 

$HERest -A -D -T 1 -C config -I aligned.mlf -t 250.0 150.0 3000.0 -S train.scp -H hmm7/macros -H hmm7/hmmdefs -M hmm8 monophones1

The files created by this command are:

 

$HERest -A -D -T 1 -C config -I aligned.mlf -t 250.0 150.0 3000.0 -S train.scp -H hmm8/macros -H hmm8/hmmdefs -M hmm9 monophones1

The files created by this command are:

Note: the monophone models created in hmm9 could actually be used with Julius for speech recognition, but recognition accuracy can be greatly improved by using Tied-State triphones - see next sections.

 


Comments

Click the 'Add' link to add a comment to this page; click the 'Read More' link to view replies to a posted comment.

AddSearch

HVite Error
By Philip - 3/10/2010

As i executed HVite in step 8 , the error is as follow


??? Error using ==> mtimes
Inner matrix dimensions must agree.

HVite Error
By Saran - 2/15/2010 - 1 Replies

While executing the HVite command of step 8. I get the following error

ERROR [+5010] InitSource : Can not open source file sp

 

 

Error HERest in Step 8
By Tronz - 1/30/2010 - 1 Replies

Hi


I receive the following error message when attempting to create hmm8 files using HERest command

 

Here is the output from HERest:

 

Processing Data: sample1.mfc; Label sample1.lab

 Error [+7332]  CreateInsts:  Cannot have Tee models at start or end of transcription

 

 

Can anyone tell me what is going wrong here an what I need to change in order to proceed?


Thanks

 

error 8050
By puphe_88 - 1/1/2010 - 1 Replies

when i running

hvite -C setting/conf2.cfg -H hmm/hmm7/macros -H hmm/hmm8/hmmdefs -S data/train/train.scp -w setting/wdnet -p 0.0 -s 5.0 -l * -y lab -m -i data/test/result.mlf setting/dict.dict setting/monophones1

i got  error +8050

The dictionary file is not correctly formatted. Section 12.7 describes the HTK dictionary file format.

 

contain of my dict.dict 

END sil

START sil

KADO k a d o

END sil

START sil

BELA b e l a

 

contain of my monophones1

a

i

ee

e

o

b

p

d

t

g

k

f

z

 

si

kh

h

j

c

m

n

ny

ng

r

l

w

y

ai

au

ou

sil

kado

bela

sp

 

contain of my grammar.gra

$bebas = BELA | KADO;

(START(<$bebas>) END)

 

where the errors occured??

 

Is not necessary a file tiedlist to use the files generate until now???
By PROYECTOSI - 12/9/2009 - 1 Replies

According to the Step 8:

Note: the monophone models created in hmm9 could actually be used with Julius for speech recognition, but recognition accuracy can be greatly improved by using Tied-State triphones - see next sections.

 

But Julius needs a file called tiedlist, when is it generated???

Is it SENT -END or SENT-END in Step 8?
By RedCisc - 7/3/2009 - 2 Replies

I get either +3219 Bad Switch END; must be single letter or +8220 LatticeFromLabels: Word SENT-END not defined in dictionary

Realignment should be run without pruning
By tpavelka - 4/23/2009 - 5 Replies

It may happen (usually during a big mismatch between transcript and the actual speech in the recording) that no tokens reach the end of the utterence due to pruning. In that case the sentence is not included in aligned.mlf, but HTK does not report any errors. When HERest is run again, it throws an error and ends because it cannot find the transcription of the sentence in aligned.mlf.

This can be solved by switching off pruning: leave out the -t switch.

Problem while re-aligning
By vkb - 10/26/2008 - 1 Replies I get the following error while re-aligning using HVite command:

ERROR [+6510] LOpen: Unable to open label file 10.lab
FATAL ERROR - Terminating program HVite

although the previous commands (HERest) worked fine without any explicit .lab files.

Please tell me where the problem is.

No sp state recognized by the HMM7
By bcbb - 12/9/2007 - 1 Replies

Hi,

There is a big problem with my trained hmms.

When applying the Hvite, I found there is no "sp" between two words.

I did sure that I confirm every step with its correctness, included the monophones1, pronunciation dictionary, etc.

 

Thanks a lot