Lower order Pade parameters miss in mglsadf

Brought to you by: s_sako, takenori-y, tokuda, uratec

Lower order Pade parameters miss in mglsadf

Forum: Open Discussion

Creator: Xingyu Na

Created: 2010-09-03

Updated: 2013-05-20

Xingyu Na - 2010-09-03

Hi,

The 2nd and 3rd order Pade parameters are missing in mglsadf, aren't they worth trying?
I calculated some according to the Pade formula, but they don't work fine.
Is there a reference?

Thanks.
Na Xingyu

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Akira Tamamori - 2010-09-03

Hi, naxingyu

The only 4th and 5th order of Pade coefft. are supported by SPTK.
The 1st, 2nd and 3rd order are not supported for now.
This is not a bug because it is known that the approximation precision in the lower order is very low.

The one of the reference of MLSA filter is;
Keiichi Tokuda, Takao Kobayashi and Satoshi Imai, ``Adaptive cepstral
analysis of speech,'' IEEE Trans. Speech and Audio Proces., vol.SA-3,
no.6, pp.481-489, Nov. 1995.

Best Regards,
Akira

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Xingyu Na - 2010-09-03

Hi Akira

Thanks very much.
I tested some lower order Pade coefft. and got resulting waveforms. It sound like some blust noises inserted into the speech.
Maybe it is because of the low approximation precision as you said.

Best regards,
Na Xingyu

If you would like to refer to this comment somewhere else in this project, copy and paste the following link:

Log in to post a comment.