Abstract

The primary function of human breathing is vital, but it also has the secondary function of facilitating speaking. During speaking, the air contained in the lungs is exhaled slowly in a coordinated manner to enable self-sustained oscillation of the vocal folds (phonation aka voicing), and articulation. The lungs are refilled by inhalation in short speaking pauses, the timing of which is organized by the speaker in a mostly unconcious way, but normally at linguistic boundaries, to ensure relative fluency of speaking despite the insertion of inhalation pauses. In a subset of speakers, exhalation happens faster than in other speakers because, e.g., vocal fold closure is impeded, resulting in excess exhalation rates (l/s) and consequently the need for more frequent inhalation pauses. Aim of this thesis is the automatic detection of inhalation events, statistical modelling of inhalation timing patterns, and anomaly detection.

Your Tasks

• literature review,
• annotate available audio recordings of text readings of speaking impaired individuals and healthy speakers wrt. inhalation events,
• train and test automatic inhale event detectors, e.g., feature based, and the wav2vec foundation model,
• analyse breathing patterns of healthy and pathological speakers wrt. the timing of inhalation
• identify speakers with abnormal breathing patterns by statistical modelling,
• documentation of the work (thesis writing, optional: paper writing)

Your Profile

• interest in speech science and technology,
• interest in health-related applications,
• good knowlegde of Python and relevant packages,
• good communication skills.

Contact

Philipp Aichinger (philipp.aichinger@meduniwien.ac.at) Barbara Schuppler (b.schuppler@tugraz.at)