@InProceedings{Sering2025_1229,
author = {Konstantin Sering},
booktitle = {Studientexte zur Sprachkommunikation: Elektronische Sprachsignalverarbeitung 2025},
title = {Smiling PAULE},
year = {2025},
editor = {Sven Grawunder},
month = mar,
pages = {3--10},
publisher = {TUDpress, Dresden},
abstract = {With the Predictive Articulatory speech synthesis Utilizing Lexical Embeddings (PAULE) model it is possible to synthesize smiled speech by spreading the lips and raising the larynx. For the articulatory synthesis of 15 German
sentences the VocalTractLab articulatory speech synthesizer is used. Synthesized
audios are evaluated in a rating experiment where 32 listeners rate the sentences
in four conditions according to the perceived smiledness of the speaker. Smiled
speech from PAULE is compared to neutral speech from PAULE and smiled and
neutral speech from Stone et al. [1]. Findings from [1] are replicated. This work
shows that the articulators in an articulatory speech synthesis model can easily be
restricted by the experimenter and that a goal directed adaptive model like PAULE
can optimize the remaining articulators to match a target acoustic as closely as possible under these restrictions.},
isbn = {978-3-95908-803-9},
issn = {0940-6832},
keywords = {Multimodal Perception of Speech and Non-verbal Cues},
url = {https://www.essv.de/pdf/2025_3_10.pdf},
}