Cumhur Erkut, Dr. Sc. (Tech)

Cumhur Erkut, Dr. Sc. (Tech)

I work on sonic and embodied interaction: model-based, neural sound and motion synthesis. Currently I am looking into foundational models and MLOps from an academic point of view.

Posts by Collection

portfolio

ImprovAIze

Published:

An AAU AI Bridging project that combines machine learning and wearable sensing devices to develop intuitive audiovisual displays that accurately reflect physical activity and the felt experience of human movement.
Read more

publications

Generative Choreographies: The Performance Dramaturgy of the Machine

Published in 15th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, Valletta, Malta, 2020

Movement generation driven by the real-time MoCap sensor data Read more

Recommended citation: Esbern Torgard Kaspersen, David Gzórny, Cumhur Erkut, and George Palamas, G. (2020). Proc. Intl. Joint Conf. Computer Vision, Imaging and Computer Graphics Theory and Applications - Volume 1, 319–326. http://doi.org/10.5220/0008990403190326

Reflections from five years of SIVE workshops

Published in J. New Music Research, 2020

Highlighted the importance of movement-based sonic interaction design and interactive machine learning in VR Read more

Recommended citation: Stefania Serafin, Cumhur Erkut, Amalia De Goetzen, Niels Christian Nilsson, Rolf Nordahl, Francesco Grani, Federico Avanzini, and Michele Geronazzo. Reflections from five years of Sonic Interaction in Virtual Environments (SIVE) workshops. 2020 J. New Music Research, 49 (1), pp 24-34 https://doi.org/10.1080/09298215.2019.1708413

A Differentiable Neural Network Approach To Parameter Estimation Of Reverberation

Published in Proc. Sound and Music Computing Conf. (France), 2022

Deep-learning real-time feedback delay network reverb as a VST3 using JUCE with CI/CD … Read more

Recommended citation: Søren V K Lyster and Cumhur Erkut, 2022. A Differentiable Neural Network Approach To Parameter Estimation Of Reverberation. In Proc. Sound and Music Computing Conf., p. 354-360, doi:10.5281/zenodo.65733571 https://doi.org/10.5281/zenodo.65733571

Pruning Deep Neural Network Models of Guitar Distortion Effect

Published in IEEE/ACM Transactions on Audio, Speech and Language Processing, 31(99), 256–264, 2022

Pruning most of the deep learning model parameters may improve the sound quality … Read more

Recommended citation: Südholt, David, Alec Wright, Cumhur Erkut, and Vesa Välimäki. 2022. “Pruning Deep Neural Network Models of Guitar Distortion Effects.” IEEE/ACM Transactions on Audio, Speech, and Language Processing 31: 256–64. https://doi.org/10.1109/taslp.2022.3223257

Differentiable All-pass Filters for Phase Response Estimation and Automatic Signal Alignment

Published in Proc DAFx 2023, Copenhagen, Denmark, 2023

learnable allpass filters optimized via an overparameterized BiasNet network without input audio. … Read more

Recommended citation: Anders Bargum, Stefania Serafin, Cumhur Erkut, and Julian D Parker. 2023. “Differentiable All-pass Filters for Phase Response Estimation and Automatic Signal Alignment.” in Proc DAFx 2023, Copenhagen, Denmark https://arxiv.org/abs/2306.00860

Reimagining Speech: a Scoping Review of Deep Learning-Powered Voice Conversion

Published in Frontiers in Signal Processin (in review), 2024

Based on scoping review of 100+ papers we outline the best practices of DL-VC priot to transformers. Read more

Recommended citation: Bargum, Anders R, Stefania Serafin, and Cumhur Erkut. 2023. “Reimagining Speech: a Scoping Review of Deep Learning-Powered Voice Conversion.” CoRR. doi:10.48550/arxiv.2311.08104. In Review: Frontiers Signal Processing https://arxiv.org/abs/2311.08104

talks

teaching