Cover photo
Gilles Degottex


Gilles Degottex

Statistical parametric synthesis  - 
Harmonic Model + Phase Distortion (HMPD) vocoder freely and openly available in COVAREP v1.3.1
with the article:

Quality similar to STRAIGHT  without voicing decision!

Have fun!
Nickolay Shmyrev's profile photoGilles Degottex's profile photo
("mixed source" like singing voice + background sounds (instruments, babbling), I assume? Not deterministic+noise glottal source.)

I've never tried. If you capture the f0 of the voice in the mixture, the harmonic synthesis will definitely reproduce a voice-like sound by generating a single harmonic structure. This should drop all the noise between the harmonics (I think this is a known result in speech enhancement). However, the harmonic structure of the voice is still disturbed by the background noise. This masks a bit the voice and regenerates part of the noise.

To investigate !
Add a comment...

Gilles Degottex

Events / Conferences  - 
The 1st ISCA Summer School: 4-8 August, Heraklion Crete Greece
The Speech Processing Courses in Crete (SPCC) are targeting to teach graduate students and researchers the latest advancements of speech processing covering theory, hands on, and establishing contacts between the academics and industry. The school will provide the chance to students and ...
Add a comment...
Basic Information
Other profiles
Contributor to