Third ESCA/COCOSDA Workshop on Speech Synthesis

November 26-29, 1998
Jenolan Caves House, Blue Mountains, NSW, Australia

Implications of Energy Declination for Speech Synthesis

Jürgen Trouvain (1), William J. Barry (1), Claus Nielsen (2), Ove Andersen (2)

(1) Institute of Phonetics, University of the Saarland, Saarbrücken, Germany
(2) Center for PersonKommunikation, Aalborg University, Denmark

ABSTRACT This paper examines whether observed phenomena in energy declination can be used to improve the naturalness of synthetic speech. In two production experiments different aspects of intensity fall-off within utterances are analysed including degree of stress, phrase length, phrase boundaries. Energy manipulation was carried out using diphone synthesis as a basis for generating stimuli for perception tests in English and Danish. The results of the listening experiments, in which different versions of a paragraph were ranked for naturalness indicate that amplitude differences can contribute to greater naturalness. However, it is apparent that fine-tuning of amplitude requires good quality synthesis at the more basic prosodic levels.

Full Paper

Bibliographic reference.  Trouvain, Jürgen / Barry, William J. / Nielsen, Claus / Andersen, Ove (1998): "Implications of Energy Declination for Speech Synthesis", In SSW3-1998, 47-52.