[hts-users:02236] Question about the generation algorithm

[YounSung Park <ys.park.hts@xxxxxxxxx>]

Dear HTS maintainer,

 

I am a student researching about the HTS based singing voice synthesis.

 

I think there is something I can't understand.

 

I saw the paper about the parameter generation algorith.

 

In the paper, the parameter generation algorithm has constraint in the generated parameter O = WC.

 

But actually, if the cepstrum vector generated from one state is long, I can't find the relationship between the cepstrum sequence.

 

( I saw the plot of the 0-th and 2nd cepstrum vector and I find that there's no discontinuity between the state transition. However after the transition and 10 frame later, the generated cepstrum value is flat..!! )

 

So I think that the parameter generation algorithm has some problem if the duration length of one state is long.

 

Thank you for reading this mail.

 

Please answer my question and let me know the solution about that problem.

 

Best Regards,

Youn-Sung Park.