Prosodic and rhythmic cues are fundamental to speech perception, as they establish the temporal organization of spoken language. This project applies psychophysical approaches (mostly auditory reverse correlation) to investigate how these cues are processed by the human auditory system. The general idea is to degrade the temporal properties, or resynthesize speech stimuli with random prosodic patterns, using speech manipulation algorithms such as WORLD-STRAIGHT and measure if/how these modified sounds are understood.

Using this approach, we have uncovered the specific prosodic cues that the auditory system relies on to segment words, for instance in ambiguous contexts such as homophonic sentences (e.g., « no notion » vs. « known ocean »). Furthermore, this work also explores how listeners dynamically adapt to variations in the temporal structure of their interlocutors’ speech by computing local estimates of speech rate.

Administrative details

Collaborators: Elsa Spinelli, Fanny Meunier, Etienne Gaudrain

Agence Nationale de la Recherche, projet DRhyaDS (2023-2026) ANR-22-FRAL-0003

Selected publications and presentations:

Scientific outreach

Présentation à la Fête de la Science (2022, ENS Paris): « La voix humaine, de Jean Cocteau aux modulations de fréquence » (slides)

Démonstration du résultat de la resynthèse sur « c’est l’amie » / « c’est la mie » : démo