Speech synthesis by structured segments, using temporal decomposition and a glottal excitation

Classical speech synthesis systems either concatenate diphone-like tabulated pattems or reconstmct speech parameters according to pre-defmed mles. Both techniques show drawbacks : the fonner lacks flexibility while the lauer is highly time-consuming_ to built. We propose an intennediate technique using structured segments : segmental units are still resorted to, but they are automatically analysed in terms of a set of spectral targets, a temporal decomposition pattem and a parametric glottal excitation. Structured segments can then be handled by rules. They also supply a valuable material which can be refered to, for building gradually a reconstructive synt.'tesis system.