MON 14 Jul 2025, 4PM, KCL STRAND S1.12
Gloria Dal Santo, PhD student in Acoustics at Aalto University to give talk at KCL Strand campus as part of the MARC Seminar Series. There will be refreshments after the talk, courtesy of the NMES Research Culture Fund.
If you are unable to attend in person, you may use the following MS Teams link to attend the event virtually: Engineering Seminar — Gloria Dal Santo
Title: Using feedback delay networks to build more flexible differentiable artificial reverberation
Abstract : Differentiable Digital Signal Processing (DDSP) is an approach that integrates automatic differentiation into traditional DSP operations, enabling backpropagation of loss gradients through DSP parameters without losing domain knowledge. The perceptual performance of a trained DDSP system is inherently tied to the choice of loss function used during optimization, particularly its ability to capture the psychoacoustic characteristics of the synthesized audio.
Feedback Delay Networks (FDNs) are recursive structures widely used for modeling room acoustics, praised for their computational efficiency. Over the past decades, research has focused on developing physically informed FDNs that incorporate information such as room geometry and surface absorption. Despite these advances, designing effective FDN parameters remains challenging, often relying on heuristics and manual tuning.
In this talk, we will explore how DDSP can aid in the design of FDNs while deepening our understanding of the relationship between acoustic descriptors and DSP parameters. By using diverse loss functions, modal analysis, and comprehensive datasets, we can work towards more flexible DDSP pipelines for artificial reverberation and spatial audio applications.
Speaker : Gloria Dal Santo received the M.Sc. degree in electrical and electronic engineering from the Ecole Polytechnique Fédérale de Lausanne, Lausanne, Switzerland in 2022, during which she interned at the Audio Machine Learning team at Logitech.
She is currently working toward a Doctoral degree with the Acoustics Lab, at Aalto University, Espoo, Finland. Her research interests include artificial reverberation and audio applications of machine learning, with a focus on designing more psychoacoustically informed systems.