Dr. Fernando Villavicencio from National Institute of Informatics, Japan, will present his work 

Some applications of Voice-Transformation to Singing-Voice and perspectives for Opera Synthesis at NII on Tuesday, November 18th 2014, 15:30, in our seminar room IC01074, Inffeldgasse 16b, first floor.

Abstract: Voice Conversion aims to the transformation of a “source” speaker in order to make it perceptually identifiable as a particular “target” one. Although the main work done in this field has been focused in spoken voice there are potential application for singing-voice synthesis (e.g. for singer’s database conversion on singing synthesis systems). In this talk we will present previous work applying Voice-Conversion and related Voice-Transformation experimentation to Yamaha’s singing-voice synthesizer “VOCALID”. The works include singer’s identity conversion by means of statistical timbre-features mapping (GMM) and accurate spectral envelope modelling. Also, aiming for large vocal-quality control from modal (normal) singing, experimentation for voice-quality transformation by source/filter estimation & transformation will be presented. The talk will end introducing the work in perspective on opera-singing synthesis by a HMM approach.