hi,

when i try to use HCopy to copy a file sample1.wav(which came along with the voxforge example) to .mfc, i encounter this message:

______________________________________________________

./HCopy -A -D -T 1 -C config sample1.wav 1.mfc

HTK Configuration Parameters[11]
  Module/Tool     Parameter                  Value
#                 ENORMALISE                 FALSE
#                 NUMCEPS                       12
#                 CEPLIFTER                     22
#                 NUMCHANS                      26
#                 PREEMCOEF               0.970000
#                 USEHAMMING                  TRUE
#                 WINDOWSIZE         250000.000000
#                 SAVEWITHCRC                 TRUE
#                 SAVECOMPRESSED              TRUE
#                 TARGETRATE         100000.000000
#                 TARGETKIND                MFCC_0

  ERROR [+6313]  OpenParmChannel: cannot read HTK Header in File sample1.wav
  ERROR [+6313]  OpenAsChannel: OpenParmChannel failed
  ERROR [+6316]  OpenBuffer: OpenAsChannel failed
  ERROR [+1050]  OpenParmFile: Config parameters invalid
 FATAL ERROR - Terminating program ./HCopy

__________________________________________________

what might be the reason? i do not foresee any problems with the wav file because it is used in training the models and its mfc file is already there with the example

Cheer

--- (Edited on 6/8/2009 7:49 am [GMT-0500] by dsubbu) ---

> i do not foresee any problems with the wav

Probably you need to look closer. Your config misses the source format of the input file:

SOURCEFORMAT = WAV

--- (Edited on 6/8/2009 2:38 pm [GMT-0500] by nsh) ---

thanks a ton! that helped.

But when I try to recognise my own test data using HVite, i get the following error

 ERROR [+3231]  ProcessFile: Incompatible sample kind MFCC_0 vs MFCC_D_N_Z_0

so i tried to change my config file for HCopy(to TARGETKIND= MFCC_0_D_Z instead of the previos one) but it does not seem to work.

Is there any way to make my test data compatible with the model(i.e. convert to MFCC_0_D_Z) other than retraining the model?

Cheers!

--- (Edited on 6/8/2009 11:53 pm [GMT-0500] by dsubbu) ---

> TARGETKIND= MFCC_0_D_Z

It should be MFCC_D_N_Z_0, just be more careful and everything will work.

--- (Edited on 6/9/2009 12:20 am [GMT-0500] by nsh) ---

im sorry, that was a typo i think.. i had used MFCC_D_N_Z_0  only.

HTK Configuration Parameters[12]
  Module/Tool     Parameter                  Value
#                 SOURCEFORMAT                 WAV
#                 ENORMALISE                 FALSE
#                 NUMCEPS                       12
#                 CEPLIFTER                     22
#                 NUMCHANS                      26
#                 PREEMCOEF               0.970000
#                 USEHAMMING                  TRUE
#                 WINDOWSIZE         250000.000000
#                 SAVEWITHCRC                 TRUE
#                 SAVECOMPRESSED              TRUE
#                 TARGETRATE         100000.000000
#                 TARGETKIND          MFCC_D_N_Z_0

  ERROR [+1019]  SetConfParms: incompatible TARGETKIND=MFCC_D_N_Z_0 for coding
 FATAL ERROR - Terminating program ./HCopy

Can HCopy convert wav files to MFCC_D_N_Z_0? if yes, what am i doing wrong?( the config file has the correct TARGETKIND)

if not, how can I do it?

--- (Edited on 6/9/2009 12:43 am [GMT-0500] by dsubbu) ---

http://www.voxforge.org/home/forums/message-boards/acoustic-model-discussions/mfcc_d_n_z_0-format

--- (Edited on 6/9/2009 12:52 am [GMT-0500] by nsh) ---

>Both operations seem to be disjoint since the MFCC_0_D to

>MFCC_D_N_Z_0 is a front end process and HCompV is used to

>initialize hmm parameters.

It would seem to be that the conversion is *not* actually done in the proto (as I had written in the article referred to in nsh's post), but in the config file.  From Step 6 of the VoxForge Tutorial:

You also need a configuration file.  Create a file called 'config' in your 'voxforge/manual' directory and add the following data:

TARGETKIND = MFCC_0_D_N_Z
TARGETRATE = 100000.0
SAVECOMPRESSED = T
SAVEWITHCRC = T
WINDOWSIZE = 250000.0
USEHAMMING = T
PREEMCOEF = 0.97
NUMCHANS = 26
CEPLIFTER = 22
NUMCEPS = 12

Which is then used inHCompV:

$HCompV -A -D -T 1 -C config -f 0.01 -m -S train.scp -M hmm0 proto

Thanks for catching that!

Ken

--- (Edited on 6/9/2009 5:31 pm [GMT-0400] by kmaclean) ---

hi Ken,

i think there is a small misunderstanding there.i'll put down my situation more clearly.

The voxforge tutorial has trained models using MFCC_0_D_N_Z type mfc files. so, the files i use to test how good the model is must also be in the same format.

i have a set of .wav files in which i have spoken only digits. Now, i want to do a batch decode (not using julian as done in the tutorial) using HVite or any other capability offered by HTK.

Now, when i convert them to .mfc, I need the config file to do a .wav to MFCC_0_D_N_Z. But HCopy does not support that. My final requirement is a .mfc file from a .wav with MFCC_0_D_N_Z format. sorry i still cant figure out how HCompV helps in this case.

Thanks a lot!

--- (Edited on 6/10/2009 6:53 am [GMT-0500] by dsubbu) ---

Hi dsubbu,

> My final requirement is a .mfc file from a .wav with MFCC_0_D_N_Z format.

HCopy can't do that directly... usually the main reason you want to convert to this particular format is for training... 

I think what you are trying to do is covered in the Testing Your Acoustic Model with HTK & Julius tutorial. 

Basically you convert your audio to MFCC_0_D (see Step 4 - Coding the Data) and then specify "TARGETKIND = MFCC_0_D_N_Z" in your config (see Test Acoustic Model Using HTK) for the decode (uses HVite in this case, but should be similar process for HDecode...).

Ken

--- (Edited on 6/10/2009 12:47 pm [GMT-0400] by kmaclean) ---