🔍 Join Our Team! We have exciting open positions:

🎓 Postdoc: Data Science & AI for Speech & Audio
🎓 PhD/Postdoc: AI-based Speech & Audio
🎓 PhD: Diffusion Models for Speech Processing
🎓 3 PhD positions for Imaging & Speech

🔗 inf.uni-hamburg.de/en/inst/ab/sp/…

Our recent paper investigates SGMSE and Schrödinger bridge for speech enhancement. We achieve 3.70 in PESQ on Voicebank-Damand.
📰 arxiv.org/abs/2409.10753
🔊 sp-uhh.github.io/gen-se/

The Signal Processing group at the University of Hamburg is looking for PhD students and a postdoc in AI-based Speech and Audio Signal Processing. We have two openings available. Apply now and join our team.
inf.uni-hamburg.de/en/inst/ab/sp/…
#phdcandidate #postdocjobs

BUDDy: Single-Channel Blind Unsupervised Dereverberation with Diffusion Models

We present a fully unsupervised method for blind speech dereverberation using a diffusion prior and a parametric subband filter

Paper 📜 arxiv.org/abs/2405.04272
Audio/Code 🔊 uhh.de/sp-inf-buddy

This is what happens when you enhance an old WHAM! interview with SGMSE+ that has been trained on EARS-WHAM.

👉Further details about the new EARS dataset can be found here: sp-uhh.github.io/ears_dataset/

youtu.be/H5FiO0JxPK4?si…

𝗣𝗵𝗗 𝗽𝗼𝘀𝗶𝘁𝗶𝗼𝗻 „𝗗𝗶𝗳𝗳𝘂𝘀𝗶𝗼𝗻-𝗕𝗮𝘀𝗲𝗱 𝗗𝗲𝗲𝗽 𝗚𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝘃𝗲 𝗠𝗼𝗱𝗲𝗹𝘀 𝗳𝗼𝗿 𝗦𝗽𝗲𝗲𝗰𝗵 𝗦𝗶𝗴𝗻𝗮𝗹 𝗣𝗿𝗼𝗰𝗲𝘀𝘀𝗶𝗻𝗴“. Join us in pushing the state-of-the-art in this exciting field.

Details and on how to apply:
inf.uni-hamburg.de/en/inst/ab/sp/…

``Diffusion Models for Audio Restoration,'' Jean-Marie Lemercier, Julius Richter, Simon Welker, Eloi Moliner, Vesa V\"alim\"aki, Timo Gerkmann, ift.tt/hPW6pEG

Happy to announce five accepted papers at @ieeeICASSP #ICASSP2024. Hope to see you there!

All papers and links to arxiv preprints can be found on our webpages
inf.uni-hamburg.de/en/inst/ab/sp/…

We just opened a PhD or Postdoc position on AI-based speech and audio signal processing.

Please see our webpage for details and on how to apply

inf.uni-hamburg.de/en/inst/ab/sp/…

It had been a super nice ICASSP in Rhodes. We really enjoyed exchanging ideas with our peers. There are so many exciting things discussed in the community atm (diffusion, self-supervised, metrics, ...). Very much looking forward for the next events to come!

Happy to announce our two recently accepted IEEE/ACM TASL journals that show how score-based / diffusion models can be used for high quality signal improvement. Audio examples, paper and code are available here
inf.uni-hamburg.de/en/inst/ab/sp/…

Happy to announce four papers at #INTERSPEECH2023 @ISCAInterspeech
on diffusion models, SSL for emotion recognition, audio-visual speech enhancement, and dereverberation. See details on our webpages
inf.uni-hamburg.de/en/inst/ab/sp/…

We offer an exciting #PhDposition on deep multi-microphone speech enhancement. Please find details on our webpages
inf.uni-hamburg.de/en/inst/ab/sp/…

Our newly developed causal version of our recent work on diffusion-based generative models for speech enhancement arxiv.org/abs/2208.05830
has been ranked in the Top 5 of the #ICASSP2023 Speech Signal Improvement Challenge and will be presented @ieeeICASSP
microsoft.com/en-us/research…

We will be presenting 7 papers @ieeeICASSP, on diffusion models, uncertainty estimation, and nonlinear joint spatial-spectral filtering, including our Top 5 contribution to the #ICASSP2023 Speech Signal Improvement Challenge
inf.uni-hamburg.de/en/inst/ab/sp/…

Happy to announce 5 accepted papers for ICASSP 2023! They address in particular
* Diffusion models
* Nonlinear joint spatial-spectral filtering
* Uncertainty Estimation in Speech Enhancement

inf.uni-hamburg.de/en/inst/ab/sp/…

#ICASSP2023

We are pleased to share our new work: „Speech Enhancement and Dereverberation with Diffusion-based Generative Models“ with exciting audio examples

Pre-print: arxiv.org/abs/2208.05830
Audio examples: bit.ly/3Ai8pqT
Code: github.com/sp-uhh/sgmse

#diffusion

Great news: We receive the VDE ITG award 2022 for our contribution which motivates the use of DNNs for nonlinear joint spatial-spectral filtering to replace sequential linear spatial filtering + postfiltering in speech enhancement. Don't miss this paper
uhh.de/inf-sp-nonline…

We are happy to announce 5 papers at Interspeech 2022 on speech enhancement and emotion recognition using e.g. Score Based Models, Bayesian DNNs, Transformers, End-to-End, and Deep Kalman
See our webpages for details uhh.de/inf-sp
#interspeech2022

Happy to announce 3 papers at #icassp2022 Customizable End-To-End Optimization Of Online Neural Network-Supported Dereverberation For Hearing Devices
Integrating Statistical Uncertainty Into Neural Network-Based Speech Enhancement
Deep Iterative Phase Retrieval For Ptychography