Multi-modal signal and image processing for speech science

In collaboration with linguists and speech scientists, the Laboratoire de traitement de l’information en santé (LATIS) has developed tools to facilitate the study of articulatory movements (tongue, lips, etc.) involved in speech production in healthy adult speakers, children, speakers with articulatory impairments or second language learners.

Several new funded Masters thesis projects are currently offered at the LATIS in this context. These projects will aim at developing various building blocks towards a framework allowing the multi- modal recording and analysis of articulatory motion during speech. The modalities involved include ultrasound imaging of the tongue, video recordings of the lips and audio recordings of speech itself. Students will contribute to the development of (1) statistical models (including machine learning models) to characterize tongue shape and motion based on tongue contours automatically extracted from ultrasound images using our pre-existing tools; (2) automated tools for tracking the inner contour of the moving lips in video sequences, accounting for variability in the appearance of individuals; (3) automated tools to align two or more audio speech signals from similar utterances (either between or within speakers) over the time dimension, which would allow meaningful statistical analysis of articulator configurations over multiple utterances or speakers. 
 

Required knowledge

Curious and self-motivated candidates interested in interdisciplinary research are sought for these projects.  Students with a computer science or engineering background should have good programming skills.  Knowledge or experience in image processing and/or computer vision and/or signal processing would be assets.  Applications from students with a linguistics/speech science/language education will also be considered. In this case, knowledge in phonetics and in quantitative methods are required.  Basic programming skills (e.g. in Python) would be an asset.

Desired program of studies

Masters with project, Masters with thesis

Research domains

Health Technologies, Sensors, Networks and Connectivity

Financing

Contact the professeur

Additional information

The LATIS is a welcoming and inclusive research environment whose success depends, among various factors, on the diversity of its members.  Therefore, applications from women, indigenous students, students belonging to visible minorities, students with disabilities or representing neurodiversity, students of all origins and all sexual orientations are highly encouraged.
 

Starting: ASAP