PhD Public Defense
Optimal Transport in Audio Signal Processing
Anton Selitskiy
Supervised by Mark Bocko
Thursday, April 16, 2026
2 p.m.–3 p.m.
426 Computer Studies Building
This dissertation investigates the role of optimal transport (OT) in modern audio signal processing and generative modeling. Many tasks in speech and audio processing— including speech synthesis, voice conversion, and domain adaptation — can be naturally formulated as transformations between probability distributions. Optimal transport provides a principled mathematical framework for studying such transformations and offers both theoretical insight and practical algorithms for modern generative models. This work combines theoretical analysis of OT with applications to speech generation, voice conversion, and adversarial analysis of audio systems.