4th International Conference on Spoken Language Processing
Philadelphia, PA, USA
A method for generating acoustic prosody is presented that starts from a very simple symbolic input. We present evidence that prominence is a central factor influencing both perception and acoustic parameters. Results of statistical analysis of a large speech corpus are shown, these results have led to the development of a rule system that predicts fundamental frequency and syllable duration. Besides the prominence of syllables and boundaries, position, context and syllable structure are considered by these rules. Finally, the outcome of two evaluation experiments is presented.
Full Paper Sound Example
Bibliographic reference. Heuft, Barbara / Portele, Thomas (1996): "Synthesizing prosody: a prominence-based approach", In ICSLP-1996, 1361-1364.