Organisiert von: Eric Starke & Jens Prager

Organisiert von: Gert Herold & Stefan Jacob

Organisiert von: Heiko Atzrodt & Felix Langfeldt

Organisiert von: Mario Wolf, Bernd Köhler & Roland Brunner 

Organisiert von: Janina Fels & Carolin Breuer

Organisiert von: Jürgen Peissig & Stefan Weinzierl

Organisiert von: Astrid Pieringer, Haike Brick, Uwe Ritterstaedt & Michael Jäcker-Cüppers

Organisiert von: Henri Siller & Martin Ochmann

Organisiert von: Claudia Lenk & Martin Ziegler

Organisiert von: Łukasz Ambroziński & Piotr Kijanka

Organisiert von: Jakob Abeßer, Sascha Grollmisch & Joachim Bös 

This session explores the expanding role of AI in environmental and industrial sound analysis. It highlights the application of machine listening techniques—such as sound event detection and localization, acoustic scene classification, and anomaly detection—in emerging scenarios, including acoustic smart cities, logistics monitoring, bioacoustics, and industrial sound classification for processes like welding and milling.
Submissions may include original research, system demonstrations, short tutorials, or overview presentations. The session places particular emphasis on the challenges posed by complex, real-world soundscapes, where variability and noise create significant barriers to robust interpretation.

Possible topics include:

• Domain Mismatch and Robustness due to Device or Location Mismatch: Ensuring that analysis systems perform well across varying recording devices or locations is crucial for applications such as context-aware hearing aids or mobile phones. The focus is on robust model architectures as well as techniques to measure and compensate for domain shift using domain adaptation.
• Domain-Specific Tasks and Challenges: Each audio domain presents its unique challenges and tasks. For example, industrial sound analysis often aims to identify very subtle changes in acoustic patterns, such as anomaly detection and failure type classification, in the presence of very loud background noise. In contrast, environmental sound analysis is faced with complex soundscapes that include both static and moving sound sources. Common tasks include sound event detection as well as animal species classification and counting in bioacoustics.
• Data Scarcity: Handling limited or imbalanced training data is a major challenge for AI-based machine listening algorithms. Possible topics include methods for synthetic data generation and data augmentation, learning techniques such as transfer learning and self-supervised learning, as well as novel conceptual approaches such as foundational models and multi-modal models, which include other data modalities such as vision, text, or specific sensors
• Efficient Neural Network Structures: Techniques like pruning, model distillation, and student-teacher learning are being applied to create lightweight, efficient models for resource-constrained devices.
• Model explainability: Methods to interpret decisions by black-box audio classification models, including for instance, post-hoc explainability methods such as Layerwise Relevance Propagation (LRP), interpretable model architectures, and sonification approaches.

Organisiert von: Volker Wittstock & Florian Schelle

Organisiert von: Julia Seitz & Jan Rennies-Hochmuth

Organisiert von: Sandra Dantscher & Helga Sukowski

Organisiert von: Jan Abshagen, Ivor Nissen, Gerhard Schmidt & Iwona Kochańska

Organisiert von: Thomas Geyer, Carsten Spehr & Ulf Tapken

Organisiert von: Anita Schulz, Stefan Schoder & Christoph Richter

Organisiert von: Anton Melnikov & Tilman Koch

This session invites scientists and researchers to present their innovative work in the rapidly evolving field of microacoustics and MEMS. The focus will be on the distinctive effects typical of microacoustics, including thermoviscous phenomena that significantly impact acoustic wave behavior at microscale dimensions. Advances in applications of various MEMS devices, such as microspeakers, microphones, and ultrasound transducers, can be shared and discussed, emphasizing recent breakthroughs and innovations. Additionally, new challenges in modeling these complex systems can be addressed, with a focus on both numerical and analytical approaches. This session aims to facilitate scientific exchange and collaboration, providing a platform for discussing the current state and future directions in MEMS technology. Researchers are encouraged to share their findings, gain insights, and contribute to the advancement of this exciting field.

Organisiert von: Stefan Balke & Jakob Abeßer & Meinard Müller

Music Information Retrieval (MIR) has evolved into a vibrant interdisciplinary field at the intersection of computer science, acoustics, and musicology. At its core, MIR addresses fundamental challenges in accessing, understanding, and interacting with music across diverse formats and data modalities. This session brings together researchers working on both foundational and applied aspects of MIR, with a focus on music processing, audio signal processing, machine learning, and artificial intelligence. Topics include deep learning techniques for raw audio analysis, hybrid and explainable modeling, the integration of musical domain knowledge into data-driven approaches, and the use of specialized music corpora for cross-disciplinary applications.

We particularly encourage experimental and exploratory contributions that investigate unconventional ideas, offer novel research perspectives, reveal the strenghts and limitations of machine learning methods in the music domain, or address educational aspects of MIR. Contributions are welcome from researchers in MIR, audio engineering, computational musicology, music recommendation, digital music libraries, and related areas.

Topics of interest include, but are not limited to:

* Deep learning for audio and symbolic music data
* Hybrid, explainable, and differentiable models
* Data preprocessing, augmentation, and annotation
* Structure analysis, transcription, and source separation
* Interactive and human-in-the-loop systems
* Applications in musicology, recommendation, and retrieval
* Open infrastructure, benchmarking, and reproducibility

Notes: Contributions may be submitted in English or German. The session will be held in English by default, with German-language presentations possible upon request.

Organisiert von: Jan-Mark Batke & Christoph Reuter 

Organisiert von: Monika Gatt & Michael Haverkamp

Organisiert von: Adam Pilch & Bartłomiej Chojnacki 

Organisiert von: Andreas Nicht & Christoph Fritzsche

Organisiert von: Martin Schneider, Adrian Blödt & Andreas Rabold

Organisiert von: Mattias Trimpop

Organisiert von: Christian Kupsch & Richard Nauber

Organisiert von: Moritz Schuck, Josep Llorca-Bofí & André Fiebig

Organisiert von: Jędrzej Kociński & Anna Warzybok

Organisiert von: Jesko Verhey

Organisiert von: Lucyna Leniowska & Dariusz Bismor