® Vocal Tract Shaping

Research and Development

  • Title: Vocal Tract Shaping
  • Real Title: The Non-Linear Dynamic Frequency Warping of the Spectral Envelope
  • Autor: Greg Beller
  • Signal Processing, Speech Modification
  • Development made in 2008

The Non-Linear Dynamic Frequency Warping of the Spectral Envelop is a signal processing technic developed during my PhD to allow for Articulation Degree alteration. The underneath basic idea is to digitally shape the vocal tract of an existing recordings to change the speaker identity or the way the speaker has pronounced an utterance. Nice side effects arise from this algorithm such as nazalisation or speaker conversion. This successful development has been afterwards hard coded into the IRCAM SuperVP engine and in the FLUX plugins IRCAM Trax V3.

Here is an example on how it sounds:

The original source utterance to be transformed…

To the high…

To the low…

To the nazalisation…

And a few pics to understand what’s happening !

MATLAB Handle Graphics


Refer to PhD dissertation and on the articles about the articulation degree