[hts-users:00195] Re: diffrences between merge and cat

["Heiga ZEN (Byung Ha CHUN)" <zen@xxxxxxxxxxxxxxxx>]

Hi,

liulei_198216@xxx wrote:

> 1.HTS voice which contains hts_engine  is  registered  by 
> \festvox\cstr_us_ked_timit_hts.scm
>  that file tells festival where is the hts voice
>  for example  the following command change the current voice :
>  festival > (voice_cstr_us_ked_timit_hts)

Yes.

> 2.\festvox\cstr_us_ked_timit_hts.scm  loads \hts\hts.scm ,
>  \hts\hts.scm calls  \hts\htsvoice.pl which uses hts_engine to 
> synthesize speech.

Yes.

> 3.\hts\hts.scm also defines some variables by which we can modify speech 
> feature,  such as f0.

Yes.

>  In the \hts\htsvoice.pl , I have found that hts_engine synthesizes speech
>  using .pdf   fles,
>  does the hts_engine have the function of generating parameters from 
> .pdf files or  the .pdf files are already the parameters format which 
> can be used to synthesize speech.

In .pdf files, statistics of HMMs are stored.
For example, mcep.pdf contains statistics of mel-cepstrum stream of HMMs.
After loading these pdf files and decision trees (.inf files), hts_engine composes a sentence HMM corresponding to a 
given label sequence.
Then speech parameter sequences (mel-cepstrum and f0 sequences) are generated from the sentence HMM using the speech 
parameter generation algorithm (please see mlpg.c in hts_engine release).
Finally, a speech waveform is synthesized from the generated speech parameter sequences using the MLSA filter.

>  In the \festvox director
>  there are some other files except cstr_us_ked_timit_hts.scm  such as
>
>  cstr_us_ked_timit_hts_phoneset.scm
>  cstr_us_ked_timit_hts_tokenizer.scm
>  cstr_us_ked_timit_hts_tagger.scm
>  cstr_us_ked_timit_hts_lexicon.scm
>  cstr_us_ked_timit_hts_phrasing.scm
>  cstr_us_ked_timit_hts_intonation.scm
>  cstr_us_ked_timit_hts_duration.scm
>  cstr_us_ked_timit_hts_f0model.scm
>  cstr_us_ked_timit_hts_other.scm
>
>  I have not found their effects in the process of synthesizing speech.
>  Why the demo of "HTS voices for fetival"  provides them?

They are provided for defining text processing conditions (e.g., POS, phone set, etc.)

Regards,

Heiga ZEN (Byung Ha CHUN)

--
 ------------------------------------------------
  Heiga ZEN     (in Japanese pronunciation)
  Byung Ha CHUN (in Korean pronunciation)

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