Inversion of F/sub 0/ model for natural-sounding speech synthesis
Pierluigi Salvo Rossi, F. Palmieri, Francesco Cutugno · 2003 IEEE International Conference on Acoustics, Speech, and Signal Processing, 2003. Proceedings. (ICASSP '03). · 2003
Natural-sounding speech synthesizers require information from a model quantitatively describing prosody. H. Fujisaki's model (see "Dynamic Characteristics of Voice Fundamental Frequency in Speech and Singing", The Production of Speech, Springer-Verlag, p.39-47, 1983) has shown considerable accuracy on many languages (Fujisaki et al., IEEE Int. Conf. on Acoustics, Speech and Sig. Processing, vol.2, p.211-14, 1993; Fujisaki and Ohno, S., Fourth Int. Conf. on Sig. Processing, vol.1, p.714-17,1998). We propose a method for the estimation of Fujisaki's model parameters, i.e., inversion methods, based on the relative extremes of the pitch contour and a gradient algorithm refinement procedure. Preliminary results show excellent performance of the proposed method in matching the pitch contours. Preliminary results of synthesis making use of the obtained features are very encouraging.