[hts-users:00509] Re: Relation between frames and MCEP order

[Omer Moav <smoavo@xxxxxxxxxxxxxxxxx>]

Our mistake,

We made a program which converts the .mcep and .lf0 to ascii file and there was
a little bug which caused it to duplicate the last line.
Now it makes sense that each frame has 25 mel-cepstral coefficients.

Thanks a lot,
Omer & Gil


Quoting "Heiga ZEN (Byung Ha CHUN)" <zen@xxxxxxxxxxxxxxxx>:

> Hi
> 
> Omer Moav wrote:
> 
> > The number of MCEP's isn't divided by 25 either.
> > We checked the number of elements in the corresponding .lf0 file (This
> should be
> > the number of frames, Isn't it?) and arrived to the following relation:
> > number_of_MCEP=(mel-cepstral_order+1)*(number_of_lf0_elements-1)+1.
> 
> Which version of HTS are you using?
> And what program do you use to generate these .mcep and .lf0 files?
> 
> > for example we got from one trial:
> > number of elements in .lf0 file: 416
> > number of corresponding MCEP's: 10376.
> > 
> > Is there a frame with only one mel-cepstral coefficient?
> > and another question: What is the default frame length in HTS?
> 
> In the demo scripts 25ms is used for frame length (5 ms is used for frame
> shift).
> 
> Regards,
> 
> Heiga ZEN (Byung Ha CHUN) 
> 
> -- 
> ------------------------------------------------
>  Heiga ZEN     (in Japanese pronunciation)
>  Byung Ha CHUN (in Korean pronunciation)
> 
>  Department of Computer Science and Engineering
>  Nagoya Institute of Technology
>  Gokiso-cho, Showa-ku, Nagoya 466-8555 Japan
> 
>  http://kt-lab.ics.nitech.ac.jp/~zen
> ------------------------------------------------
>