During this presentation, voice transformation will be discussed as well as various existing point of view of this research subject. First, some results of my Ph.D. will be presented. The SVLN method, a signal model inspired from voice production, will be described (Separation of the Vocal-tract with the Liljencrants-Fant model + Noise). Examples of voice transformation and a very short contemporary composition using this voice model will be also presented. Then, I will shortly compare voice transformation technics in order to discuss the main issues of voice modeling for voice transformation and speech synthesis. Given this context, I will finally desribe the theme of my post-doc research at UOC.
Gilles Degottex received the Diploma degree in computer science in 2003 from University of Neuchatel, Switzerland. After a one-year specialization in signal processing at Ecole Polytechnique Federale de Lausanne, Switzerland, he obtained the Ph.D. degree (summa cum laude) at the Institut de Recherche et Coordination Acoustique/Musique (Ircam), Paris VI, in the Analysis/Synthesis Team. His research interests include fundamental frequency tracking for musical instruments, parameter estimation of glottal model and voice modeling for voice transformation and speech synthesis.
