[hts-users:01248] Re: Questions concerning the HTS_engine

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

Hi,

Miguel Vaz wrote (2008/03/13 21:21):

> 1) I've seen one can get the cepstral features with the HCopy command.

You cannot get cepstral coefficients for speech synthesis with the HCopy command.
You can get MFCCs, but it cannot be used for speech synthesis directly.

> Is there anyway of getting the MLSA filter and using it only using the 
> available commands, or would I have to write a program myself using the 
> HTS_engine API?

Use SPTK.

> 2) Supposing I pass point 1, is it possible to choose between a mixed 
> excitation model and a simple pulse one?

Currently only simple excitation model is provided in HTS and SPTK.

> 3) Is there any work comparing the quality of the synthesized speech 
> using different parameters (number of coefficients, etc)? 

Internally we did a number of experiments and one of them has been published in a domestic conference.

> Is there any optimal configuration?

These parameters are usually speaker-dependent, 
so I think there is no optimal configuration. 

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