[hts-users:04133] Re: Dealing with "metallic" sounding of some consonants in HTS

[Ilya Edrenkin <ilia@xxxxxxxxxxxxxxx>]

Hi Momop,

 

Sure: I've changed it in the synthesis parameters (-u parameter if you use the hts_engine API). Probably I should have called it "voiced/unvoiced threshold" to be clear.

 

Ilya

 

02.10.2014, 05:53, "Momop Momop" <momop540@xxxxxxxxx>:

 

Illya

 

Can you please tell me where you changed the MSD threshold ?

 

Thank you

momop

On Wednesday, October 1, 2014 1:18 PM, Ilya Edrenkin <ilia@xxxxxxxxxxxxxxx> wrote:

Hi,

 

Has anyone faced a problem of "metallic" or "ringing" sounding of particular consonants in HTS, such as  "z" and "th"?

 

Here are a couple of examples attached: one English "this is the zombie" from cmu-slt voice and one Russian with text "zoya zabrala zebru". In both cases "z" sounds a bit strange in almost the same way, although the training databases are disjoint and even the training systems are different (English is taken from MaryTTS, Russian is built on hts-2.3alpha).

 

Training data itself does not contain such an effect for "z" phones. Postfiltering or altering GV weights doesn't seem to help. Setting MSD threshold as high as 0.95 does help: there is no more "metallic" sounding for "z", but vowels are of course distorted into "whispering".

 

It seems that the main problem is in the frequency band 2.5KHz-6KHz. It is visible at the spectrogram (attached; positions 0.2, 0.5, 1.0).  Applying a steep bandreject filter (almost removing this band) does help. I wonder if there is a cleaner way to deal with it? Probably tuning the MLSA filter or playing with mel-cepstral feature extraction parameters could help?

 

Thank you for any advice!

 

Regards,

Ilya