[hts-users:01306] Re: Questions of HMGens

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

Hi,

孙敬伟 wrote (2008/04/03 11:07):

>   In option '-c', it is "1: mix sequence is given, but state sequence is 
> hidden", but I think it should be "state sequence is given, mix sequence 
> is hidden", right?

Which version are you using?
I think it has been fixed since version 2.0.1.

>   In options '-e', '-m', '-s', what does "align" mean here,  what is the 
> true meaning of these options, could you  please explain them more 
> explicitly?

"-e" uses phoneme boundaries from label files to prune forward-backward probability calculation.
"-m" uses phoneme boundaries from label files to set phoneme durations of synthesized speech.
"-s" uses state-level boundaries from label files to set state durations of synthesized speech.

>   In options '-E', '-J', what does "xform" mean?

Linear transformations for adaptations.
These options are also used in HERest, please check HTKBook.

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://www.sp.nitech.ac.jp/~zen
------------------------------------------------