[Subject Prev][Subject Next][Thread Prev][Thread Next][Date Index][Thread Index]

[hts-users:00536] Duration model



Hi,Heiga
In HTS 2.0,duration is modeled by Gaussion distribution.Now I am attempting to model it with Gamma distribution.
I read the source code according to the paper,"Duration Modeling For HMM-Based Speech Synthesis".However I couldn't find out how the duration is modeled by Gaussion distribution.It seems it doesn't matter much that which distribution is used to model the duration.
 
Best Regards!
Lifang 
"Heiga ZEN (Byung Ha CHUN)" <zen@xxxxxxxxxxxxxxxx> 写道:
Hi,

Fang Li wrote:

> In HTS-2.0,HRest can output duration models.In the duration model file,I
> saw that the DurKind is <.GEND>. What does mean? Does it mean
> general model?

I don't know.
, , , , and have been defined by HTK.
Only was added by HTS.
No information about DurKind is provided in HTKBook.

# We can guess the meaning of POSSIOND and GAMMAD.

> HERest can also output duration models when the option -g was set. This
> invoke a function SaveDuration(fbInfo,durFN,GAUSSD). I believed it means
> a gaussian model of duration will be saved. Then I looked at the source
> code of SaveDuration.I found this line:
> fprintf(fp,"~o\n %d\n",MaxN);
> , rather than , appears again!
> Do they have the same meaning?

Currently DurKind is not used in any HTK/HTS routines.
So changing to has no effect.

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



抢注雅虎免费邮箱-3.5G容量,20M附件!
Follow-Ups
[hts-users:00537] Re: Duration model, Heiga ZEN (Byung Ha CHUN)
References
[hts-users:00502] Re: and Duration model, Heiga ZEN (Byung Ha CHUN)