In order to account»for the role of focus words, i. e. words bearing the highest load of semantic information in speech, we introduced the so-called 'peak-and-level' model. By means of this model we studied acoustic-phonetic characteristics of focus words in spontaneous speech as well as in the same texts, read aloud after orthographic transcription. The results of our measurements in both natural speech conditions were compared with the data when the same texts are synthesized by Dutch diphone synthesis. After manipulation of some spectro-temporal aspects in the speech synthesis, listeners were asked to judge the naturalness and intelligibility of the speech, as compared to the original spontaneous and read versions.
Bibliographic reference. Beinum, Florien J. Koopmans-van (1991): "A peak-and-level model for focus words in read and spontaneous natural speech and in synthetic speech", In EUROSPEECH-1991, 1459-1462.