[hts-users:00708] Re: HTS for separate pitch training

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

Hi,

vancy wrote (2007/06/08 16:00):

> I'm trying to do some work about HMM-based TTS, and want to train HMM 
> models for f0 contour. I have a few questions about HTS.
>
> 1. Whether HTS can train a f0 model separately without spectrum and 
> duration feature in the uniform HMM framework as it mentioned in the paper.

You can train f0 models separately.
However, sometimes f0 features are not enough to obtain appropriate phoneme assignments.

For example, if a series of unvoiced phones appears, we may observe a sequence of unvoiced symbols.
In this unvoiced region, we cannot get any information from the observations to determine phoneme assignments.
So state transitions in this region may be corrupted :-(

> 2. Is there any user guider for HTS training like HTKBook? or only to 
> investigate the code of revised HTK Tools?

Currently no user guide for HTS training is available.
However, you can use this mailing list and its archive (http://hts.sp.nitech.ac.jp/?Mailing%20List).

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