[hts-users:00019] Re: ESPS get_f0

[Heiga ZEN <zen@xxxxxxxxxxxxxxxx>]

Hi Nicholas,

On Sun, 21 Mar 2004 15:38:57 +0200 (EET)
"Nicholas Volk" <nvolk@xxxxxxxxxx> wrote:

> After an initial look at the HTS-demo_CMU-ARCTIC-AWB,
> which I try to imitate, there are few questions that puzzle me...
> 
> 1) The size of *.raw files is always divisible by 32000. Why?
> I don't think this is the case in the original Artctic files.
> Is this necessary and do you just append silence to the end of the
> files?

No, it isn't necessary.
I cut the header of wav files included in the original Arctic database by "sox" to make raw files in HTS-demo_CMU-ARCTIC-AWB.

> 2) I got the ESPS from KTH's web site, but it's notoriously hard to install
> (at least in Linux and Cygwin/Windows). 
> Does anyone know of precompiled
> packages or some other way to get ESPS's get_f0 working?

My colleague did it, but I think he have already forgotten how to compile them.
 
> Or is there a way to convert data from some other F0 extractor to
> ESPS's *.f0 files?

The *.f0 files included in HTS-demo are not ESPS's f0 format.
They are binary float files in little endian.

> For example Festival's Edinburgh Speech Tools comes with a tool called pda
> which also extracts F0 contour. It might not be as good as get_f0
> (the algorithm is different), but it would probably suffice.

If the extracted f0 file is in ascii format, 
the command "x2x" in SPTK can convert it into HTS-demo format.

x2x +af f0.ascii > f0.binary   (ascii->float)


Best regards,

Heiga Zen

-- 
 ------------------------------------------------
  Heiga Zen
  Department of Computer Science and Engineering
  Graduate School of Engineering
  Nagoya Institute of Technology
  Japan

  e-mail: zen@xxxxxxxxxxxxxxxx
     web: http://kt-lab.ics.nitech.ac.jp/~zen
 ------------------------------------------------