Third International Conference on Spoken Language Processing (ICSLP 94)
This paper describes a hybrid harmonic plus noise system developed for TTS applications in the CNET multilingual concatenation based environment. It was implemented in order to introduce more naturalness and variability in the speech synthesis, enabling greater prosodic modifications than is currently generated by PSOLA type systems. Another motivation for this research was the need to access the fine structure of the speech signal in order to perform voice transformations. An evaluation prototype was tested with a complete set of diphones extracted from a french male voice and the preliminary results are very encouraging.
Bibliographic reference. Boeffard, O. / Violaro, F. (1994): "Using a hybrid model in a text-to-sppech system to enlarge prosodic modifications", In ICSLP-1994, 727-730.