This paper presents a prosodic analysis of a corpus of 12 tales, read by one male speaker. The work is part of a project which aims at providing storytelling capacities to a humanoid robot. One main point is to improve text-to-speech synthesis expressivity according to a semi-automatic analysis of a given tale. Automatic tagging and prosodic stylization were applied to the corpus. The extracted parameters are described and analyzed according to relevant elements of the tales' structure. The results show that changes of pitch and intensity registers, use of glide tones or devoicing are relevant to impersonate prototypical characters and to modify expressivity according to the different structural parts of each tale. These prosodic results will contribute to enhance the expressivity of a non-uniform-unit text-to-speech synthesizer.
Cite as: Doukhan, D., Rilliard, A., Rosset, S., Adda-Decker, M., d'Alessandro, C. (2011) Prosodic analysis of a corpus of tales. Proc. Interspeech 2011, 3129-3132, doi: 10.21437/Interspeech.2011-783
@inproceedings{doukhan11_interspeech,
author={David Doukhan and Albert Rilliard and Sophie Rosset and Martine Adda-Decker and Christophe d'Alessandro},
title={{Prosodic analysis of a corpus of tales}},
year=2011,
booktitle={Proc. Interspeech 2011},
pages={3129--3132},
doi={10.21437/Interspeech.2011-783}
}