[hts-users:00361] Re: about " decision trees"

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

Hi,

lei liu wrote:

> 1.  about  mcep and  f0 trees
> when we have used  states  to   construct  the tree,
> the leaf of the tree will represent a set of states   which  have the same
> contexts.
> But the leaf node have only one pdf which is the   distribution in that
> contexts' condition.
>
> How to determine the pdf though the set of states in the leaf node.

A set of states are merged and 1 pdf is generated.
Please read 

S.J. Young, J.J. Odell and P.C. Woodland, 
"Tree-based State Tying for High Accuracy Acoustic Modeling," 
Proc. ICASSP’94 , pp.307–311.

> 2.about the MDL-based decision tree
> Where can i get any papers about the MDL-based decision tree'
> constrution  especially about  the duration tree  ?
> The paper
> "SIMULTANEOUS MODELING OF SPECTRUM, PITCH AND DURATION IN HMMBASED SPEECH
> SYNTHESIS"
> only have a  general introduction.

K. Shinoda and T. Watanabe, 
"Acoustic modeling based on the MDL principle for speech recognition”,
Proc. EUROSPEECH'97, pp. 99-102.

T. Yoshimura, K. Tokuda, T. Masuko, T. Kobayashi, T. Kitamura,
"Duration Modeling in HMM-based Speech Synthesis System,"
Proc. ICSLP'98, pp.29-32.


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