Dynamics of vocal identity in deceptive speech
Supervisor(s): Alessandro De Luca, Volker Dellwo
Summary
In the NumberLie Game project we are collecting and analysing speech data within the context of a social game played between 5 participants, with the goal of understanding if there are differences in vocal identity (expressed as different acoustic features) between non-deceptive and deceptive speech for the same speaker.
For MA Thesis
-
Native speaker of Mandarin, Italian, or English
-
Goals: collect data using the NumberLie Game study design (https://www.isca-archive.org/interspeech_2024/deluca24_interspeech.html#), analyse and compare across languages within-speaker variability in deceptive speech OR other analyses relating to deceptive speech (open to student’s suggestions and ideas).
For Programming Projects
Programming project:
-
Goals: modelling of acoustic features related to voice identity (MFCCs, FBank, etc.) for visualisation using different reduction approaches, and in terms of geometry or graph networks in high-dimensional space.
-
Requirements: Python or other programming language.
-
Desirable requirements: interested in phonetics and speech sciences and interested in adding their own input and ideas to the project.
Requirements
MA thesis
Desirable requirements:
- Willingness to review literature independently, openness to discussions and bringing own ideas to meetings.
- Background knowledge in phonetics / signal processing.
Programming Projects
- Requirements: Python or other programming language.
- Desirable requirements: interested in phonetics and speech sciences and interested in adding their own input and ideas to the project.