[hts-users:00253] Re: about HTS-demo

["Nicholas Volk" <nvolk@xxxxxxxxxx>]

Hi,

With TCL/Snack you can write a tcl script to do the job.
I've sent to the mailing list a pointer to Snack tutorial
at http://www.speech.kth.se/snack/tutorial.html which contains
some code that can serve as example on how to use Snack to get F0.
The example "Mean pitch per speaker" doesn't do exactly what you want, but
it's not too hard to write a script that does so (based on the example) if
you know rudimentary scripting.

More generally: If you have 4 fields and you'll need but one of them,
you might use awk, perl or cut (just to name a few).

Of the F0 extraction tools I mentioned, PDA is the easiest to use:

pda -fmin $FMIN -fmax $FMAX wav/$i.wav -otype ascii |
$SPTKBINDIR/x2x +af > foo.f0

(modified from unpublished build_hts script by AWB, you may have to change
the frame size (-S option) to whatever you happen to use.)

The f0 file contains but F0 values for every "time frame" (typically 5ms
period). To convert f0 file to a human readable format,
one can use "x2x +fa".

regards,
  Nicholas

>
> Hi
> thank you , Nicholas Volk.
> I have used the method of ESPS/Snack,
> and got  the ".f0" file.
> In the ".f0" file, there are 4 fields for each  frame, pitch, probability
> of voicing, local root mean squared measurements,  and the peak normalized
> cross-correlation value.
> But  I think   the  ".f0" file    in the  "HTS-demo_NIT-ATR503-M001"
> has the different  format,
> It  maybe contains the relations between phoneme  label and f0.
> How can I get that  ".f0" file from raw  file.
> Best  regards
> liulei      From:  "Nicholas Volk"
> Reply-To:  hts-users@xxxxxxxxxxxxxxxxxxxxxxxxx
> To:  hts-users@xxxxxxxxxxxxxxxxxxxxxxxxx
> Subject:  [hts-users:00242]  Re: about HTS-demo
> Date:  Sat, 8 Apr 2006 14:18:13  +0300 (EEST)
>>Hi,
>>
>>I merely said that Praat has the  ability to calculate the pitch.
>>Nobody is going to go manually  through hundreds of files,
>>but it's  relatively easy to  write a Praat script to do the job.
>>
>>Essentially it's just a  simple loop which prints
>>the F0 value after every N milliseconds for  the whole duration
>>of the input wave.
>>My twenty line script  is proprietary code so I can't give it away.
>>
>>PDA from  Edinburgh is easier to use, but it's not too good with
>>creaky  voices.
>>
>>http://festvox.org/docs/speech_tools-1.2.0/x2152.htm
>>
>>HTS  people apparently use get_f0 from ESPS/Snack.
>>Section "Pitch and  formant extraction"  in
>>http://www.speech.kth.se/snack/tutorial.html#batch
>>for a  similar example (which can not be used as such  though).
>>
>>regards,
>>    Nicholas
>>
>> >
>> > Hi,
>> > Thank  you.
>> > I have used the Praat for get f0,
>> > the result  is ".pitch" file ,
>> > is it any difference between ".pitch" and  ".f0" file.
>> >
>> > Can I get the ".f0" through modifying  the filename extension.
>> > Best regards
>>  >      From:  "Nicholas  Volk"
>> >  Reply-To:  hts-users@xxxxxxxxxxxxxxxxxxxxxxxxx
>> >  To:  hts-users@xxxxxxxxxxxxxxxxxxxxxxxxx
>> >  Subject:  [hts-users:00240]  Re: about HTS-demo
>>  > Date:  Sat, 8 Apr 2006 12:41:52  +0300  (EEST)
>> >>The HTS mailing list archives might contains  some  clues...
>> >>
>> >>For F0 one might  use get_f0 (in Snack), pda (in  Edinburgh Speech
>> Tools)
>>  >>or Praat, just to name a  few.
>> >>
>>  >>For labels you might use Festival. It has support  for  English.
>> >>Also the unofficial 2.1 version of  Festvox appears to  contain
>> >>rudimentary support  for building and english  voice.
>> >>Link to it may  (or may not) be in festival/festvox mailing  list
>>  >> archives.
>> >>
>> >>
>>  >>  Nicholas
>> >>
>>  >>  > Hi
>> >> > In the HTS-demo, what  tools do you use for getting  ".f0"  file
>> and
>>  >> labling
>> >> > " .lab" file.
>> >>  >  Thank you.
>> >> > Best regardsÃâ·ÑÏÂÔØ   MSN  Explorer
>> >>
>>  >>
>> > Ãâ·Ñ ÏÂÔØ   MSN  Explorer
>>
>>
> Ãâ·ÑÏÂÔØ  MSN Explorer  Get 2 months FREE*.