Dissertation Julius Richter

12 September 2025, by David Mosteller

Photo: Gerhard Richter

Julius Richter successfully defended his PhD degree with his thesis "Generative Speech Enhancement in Multimodal Applications". Thanks to committee members and in particular the external members Shinji Watanabe and Simon Leglaive!

Julius Richter's dissertation advances generative speech enhancement by developing innovative unsupervised and supervised machine learning approaches that integrate visual information, including new Bayesian and diffusion-based models. Through extensive evaluation on a novel, high-quality speech dataset, the work demonstrates that these generative and audio-visual methods outperform existing techniques, establishing their effectiveness for robust speech enhancement and audio restoration.