[hts-users:00537] Re: Duration model

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

Hi,

Fang Li wrote:

> In HTS 2.0,duration is modeled by Gaussion distribution.Now I am 
> attempting to model it with Gamma distribution.

My colleagues have already applied Gamma distribution for state duration modeling.
It was published as

Y. Ishimatsu, K. Tokuda, T. Masuko, T. Kobayashi, and T. Kitamura,
"Investigation of state duration model based on gamma distribution for HMM-based speech synthesis,"
Technical Report of IEICE, SP2001-81, vol.101, pp.57–62, 2001. 

> 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 Gaussian distribution.

Formulas in that paper includes a fatal error.
Corrections are available at 

http://hts.ics.nitech.ac.jp/publications/dur-correct

From HTS version 1.1b, corrected version has been used.

Best 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
------------------------------------------------