Discrimination de styles de parole par analyse prosodique semi-automatique

This study focuses on prosodic differences between speaking styles, and their automatic distinction. We aim at characterizing speaking styles with the purpose of distinguishing them from each other, modeling them and eventually adding expressivity to text-to-speech systems. This can be done with a multi-level annotation of varied corpora, based on automatic processes (like phonetic segmentation, grammatical tagging) and manual annotations (of perceived syllabic prominences and of delivery speech objects). Quantitative comparisons of various prosodic parameters are conducted through the acoustic and linguistic dimension to catch differences between speaking genre