Speech Recognition Engines

Flat
Who can deal with .wav file in HTK??? Thank you so much for your help!!!!
User: NancyZxll
Date: 5/1/2012 6:45 pm
Views: 5599
Rating: 7

Hello there,

    Still, I cannot deal with my .wav files generated by Audacity in HTK. My HTK can only deal with the .sig files generated by HSlab.

    Since I need to do comparison experiment, I recorded my voice through audacity followed the recording toturials.

    I have my sample data and all the additional files when running HCopy and HCompV.

    I really need help:"( If I cannot solve the problem, I can't do my comparison experiment:"(

   Who can help me? Can I send all my sample files to anyone else? Is my HTK wrong??

   Thank you all for your great help!:"(

--- (Edited on 5/1/2012 6:45 pm [GMT-0500] by NancyZxll) ---

--- (Edited on 5/1/2012 6:46 pm [GMT-0500] by NancyZxll) ---

Re: Who can deal with .wav file in HTK??? Thank you so much for your help!!!!
User: kmaclean
Date: 5/1/2012 8:13 pm
Views: 233
Rating: 10

>I cannot deal with my .wav files generated by Audacity in HTK. My HTK

>can only deal with the .sig files generated by HSlab.

which O/S are you using? which version of HTK? were there any errors when you compiled your HTK? 

--- (Edited on 5/1/2012 9:13 pm [GMT-0400] by kmaclean) ---

Re: Who can deal with .wav file in HTK??? Thank you so much for your help!!!!
User: NancyZxll
Date: 5/2/2012 11:54 am
Views: 239
Rating: 8

Hello Kmaclean,

   I am using Windows, 32bits.

   The HTK is the "current stable release" version: 3.4.1

   I compiled my HTK according to the tutorial: http://htk.eng.cam.ac.uk/docs/inst-win.shtml

   There was no error happend when compiling, but when I test my installation, I run 

cd HTKDemo

perl runDemo.pl configs/monPlainM1S1.dcf 

I didn't see the result shown in the tutorial. My result is very long, the last several paras were:

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

 - system is PLAIN

90 Observation Sequences Loaded

Starting Estimation Process

Iteration 1: Average LogP =  -531.87140

Iteration 2: Average LogP =  -527.31677  Change =     4.55465

Iteration 3: Average LogP =  -526.60223  Change =     0.71452

Iteration 4: Average LogP =  -526.32343  Change =     0.27880

Iteration 5: Average LogP =  -526.26605  Change =     0.05737

Iteration 6: Average LogP =  -526.24451  Change =     0.02152

Iteration 7: Average LogP =  -526.23560  Change =     0.00892

Iteration 8: Average LogP =  -526.23358  Change =     0.00200

Iteration 9: Average LogP =  -526.22858  Change =     0.00502

Iteration 10: Average LogP =  -526.22864  Change =    -0.00007

Estimation converged at iteration 11

Output written to directory hmms/hmm.0

 

HTK Configuration Parameters[3]

  Module/Tool     Parameter                  Value

                  KEEPDISTINCT               FALSE

                  SAVEGLOBOPTS                TRUE

                  TARGETKIND              MFCC_E_D

 

 

 

Calling HInit for HMM V

HInit -A -i 10 -L labels\bcplabs\mon -l V -o V -C toolconfs\hinit.conf -D

s\hmm.0 -T 1 -S train.scr proto/V

 

HTK Configuration Parameters[3]

  Module/Tool     Parameter                  Value

#                 KEEPDISTINCT               FALSE

#                 SAVEGLOBOPTS                TRUE

#                 TARGETKIND              MFCC_E_D

 

Initialising  HMM proto/V . . .

 States   :   2  3  4 (width)

 Mixes  s1:   1  1  1 ( 26  )

 Num Using:   0  0  0

 Parm Kind:  MFCC_E_D

 Number of owners = 1

 SegLab   :  V

 maxIter  :  10

 epsilon  :  0.000100

 minSeg   :  3

 Updating :  Means Variances MixWeights/DProbs TransProbs

 

 - system is PLAIN

68 Observation Sequences Loaded

Starting Estimation Process

Iteration 1: Average LogP =  -599.68665

Iteration 2: Average LogP =  -592.98315  Change =     6.70348

Iteration 3: Average LogP =  -592.21436  Change =     0.76877

Iteration 4: Average LogP =  -591.48163  Change =     0.73274

Iteration 5: Average LogP =  -591.14313  Change =     0.33848

Iteration 6: Average LogP =  -591.08783  Change =     0.05529

Iteration 7: Average LogP =  -591.04993  Change =     0.03791

Iteration 8: Average LogP =  -590.98993  Change =     0.05998

Iteration 9: Average LogP =  -590.98560  Change =     0.00435

Iteration 10: Average LogP =  -590.98560  Change =     0.00001

Estimation converged at iteration 11

Output written to directory hmms/hmm.0

 

HTK Configuration Parameters[3]

  Module/Tool     Parameter                  Value

                  KEEPDISTINCT               FALSE

                  SAVEGLOBOPTS                TRUE

                  TARGETKIND              MFCC_E_D

 

 

 

Calling HInit for HMM N

HInit -A -i 10 -L labels\bcplabs\mon -l N -o N -C toolconfs\hinit.conf -D

s\hmm.0 -T 1 -S train.scr proto/N

 

HTK Configuration Parameters[3]

  Module/Tool     Parameter                  Value

#                 KEEPDISTINCT               FALSE

#                 SAVEGLOBOPTS                TRUE

#                 TARGETKIND              MFCC_E_D

 

Initialising  HMM proto/N . . .

 States   :   2  3  4 (width)

 Mixes  s1:   1  1  1 ( 26  )

 Num Using:   0  0  0

 Parm Kind:  MFCC_E_D

 Number of owners = 1

 SegLab   :  N

 maxIter  :  10

 epsilon  :  0.000100

 minSeg   :  3

 Updating :  Means Variances MixWeights/DProbs TransProbs

 

 - system is PLAIN

16 Observation Sequences Loaded

Starting Estimation Process

Iteration 1: Average LogP =  -398.18735

Iteration 2: Average LogP =  -395.48715  Change =     2.70020

Iteration 3: Average LogP =  -395.26947  Change =     0.21768

Iteration 4: Average LogP =  -395.17661  Change =     0.09286

Iteration 5: Average LogP =  -395.17661  Change =     0.00000

Estimation converged at iteration 6

Output written to directory hmms/hmm.0

 

HTK Configuration Parameters[3]

  Module/Tool     Parameter                  Value

                  KEEPDISTINCT               FALSE

                  SAVEGLOBOPTS                TRUE

                  TARGETKIND              MFCC_E_D

 

 

 

Calling HInit for HMM L

HInit -A -i 10 -L labels\bcplabs\mon -l L -o L -C toolconfs\hinit.conf -D

s\hmm.0 -T 1 -S train.scr proto/L

 

HTK Configuration Parameters[3]

  Module/Tool     Parameter                  Value

#                 KEEPDISTINCT               FALSE

#                 SAVEGLOBOPTS                TRUE

#                 TARGETKIND              MFCC_E_D

 

Initialising  HMM proto/L . . .

 States   :   2  3  4 (width)

 Mixes  s1:   1  1  1 ( 26  )

 Num Using:   0  0  0

 Parm Kind:  MFCC_E_D

 Number of owners = 1

 SegLab   :  L

 maxIter  :  10

 epsilon  :  0.000100

 minSeg   :  3

 Updating :  Means Variances MixWeights/DProbs TransProbs

 

 - system is PLAIN

25 Observation Sequences Loaded

Starting Estimation Process

Iteration 1: Average LogP =  -424.39636

Iteration 2: Average LogP =  -420.31281  Change =     4.08355

Iteration 3: Average LogP =  -419.43472  Change =     0.87808

Iteration 4: Average LogP =  -419.35437  Change =     0.08035

Iteration 5: Average LogP =  -419.32385  Change =     0.03050

Iteration 6: Average LogP =  -419.32385  Change =    -0.00001

Estimation converged at iteration 7

Output written to directory hmms/hmm.0

 

HTK Configuration Parameters[3]

  Module/Tool     Parameter                  Value

                  KEEPDISTINCT               FALSE

                  SAVEGLOBOPTS                TRUE

                  TARGETKIND              MFCC_E_D

 

Can't open hmms\hmm.1

------------------------------------------------------------------
Even though I didn't see the statistics result, I think there wasn't any error with HTK, was there?

 

--- (Edited on 5/2/2012 11:54 am [GMT-0500] by NancyZxll) ---

Re: Who can deal with .wav file in HTK??? Thank you so much for your help!!!!
User: kmaclean
Date: 5/2/2012 12:38 pm
Views: 98
Rating: 10

>I didn't see the result shown in the tutorial.

Have you seen these posts: Compiling HTK 3.4 on Windows XP 

have you tried installing HTK under cygwin?

--- (Edited on 5/2/2012 1:38 pm [GMT-0400] by kmaclean) ---

Re: Who can deal with .wav file in HTK??? Thank you so much for your help!!!!
User: NancyZxll
Date: 5/2/2012 1:48 pm
Views: 2213
Rating: 11

Hello Kmaclean,

   So, you mean the HTK can deal with .wav files under Linux, right?

I have installed Linux VMare Machine on my windows. If the HTK under Linux windows can do it successfully, I will try to install the HTK again on my Linux VMare Machine.

    Thank you so much Kmaclean!

--- (Edited on 5/2/2012 1:48 pm [GMT-0500] by Visitor) ---

PreviousNext