FILTRAGE RYTHMIQUE DES DONNÉS ACOUSTIQUES EN PAROLE SPONTANÉE : UN TRAITEMENT POUR LA FO ET LA DURÉE SEGMENTALE

Abstract
In spontaneous speech, the study of intonation postulate to call into question the stereotypes stemming from the studies of reading materials. From our point of view, the place of accent is not predictable, but it can be detected according to rhythmic cues. Fundamental frequency seems to be the essential one which a duration cue can be optionaly added. The presence/absence of duration cues enables us to build rhythmic hierarchy based upon perception processes. Duration lenghtening alone, i.e. without any f0 cue, often corresponds to a phenomenon called "vocal hesitation". In this paper, we try to show both acoustic behavior and rhythmic and perceptual functions of prosody in spontaneous speech.