Distributed Listening to the City:
a mobile phone-based creative recorder for urban sounds
Marcel Zaes Sagesser, Xuehua Fu, Sihui Li

How can we enhance our understanding of dense, large, rapidly developing megacities by
listening to them? Shenzhen City offers an example of a fast-changing urban space that, in
terms of its sonic environment, is complex, given the city comprises different strata of
technologies, classes, industries, businesses, transportation modalities, and natural features.
This project develops what we call a “distributed” system through which untrained users,
using their personal mobile phones, listen to, and record Shenzhen. This participatory
approach to understanding the city through sound allows us to collect qualitative data from
users who move in their city on an everyday basis, in addition to building an audio data base
of the city. We prompt users to add photos, tags, keywords, description, and emotional
attributes to what they record into the database.
In this talk, we introduce the unique creative method to listen to the city via a distributed
system that utilizes ubiquitous technology in the contemporary City of Shenzhen – mobile
phones. We present our app prototype that we developed and the results of a small-scale
pilot study, which points to the vast potential of this approach on shedding qualitative
knowledge on Shenzhen’s soundscape that can extend purely computational methods.
soundstudiesgroup.net

The versatility of mobile technologies in music from concerts to classrooms 
Oliver Hödl

Abstract:
This presentation provides an insight into various research projects exploring the use of
mobile technology in the field of music. Mobile technologies are utilised as sensor-based
interfaces for new digital musical instruments, tools for interactive audience participation
and embodied effects control, as well as platforms for distributed performances employing
ultrasound communication. The applications this research is embedded span music
concerts, interactive installations, and educational contexts.

Short bio:
Oliver Hödl is a computer scientist and independent musician, holding a professorship at the
University of Applied Sciences Upper Austria and teaching as a Privatdozent at the
University of Vienna. He spent significant parts of his doctorate – completed in 2016 with
distinction – in England at universities in Nottingham, Milton Keynes, and London. He has
successfully led large international research projects with total budgets of 30 million euros,
with over 24 million euros under his responsibility as Principal Investigator (PI). He has an
extensive track record of more than 50 publications, and his artistic work includes over 100
concerts and installations worldwide, more than 30 compositions, and 4 CD releases.

Borderline Physical: An Interim Report
Patrik Lechner

The talk presents preliminary findings and experiments from Lechner’s ongoing participation
in the Artistic Doctorate Programme at the University of Music and Performing Arts Vienna.
The core idea behind this research is to digitally model aspects of our reality and explore the
resulting affordances in a radical way. Simulating everyday objects—such as a fridge—
produces complex audio-visual models with countless parameters, which are then explored
as new forms of instruments for performance and composition.

As part of the research project Spirits in Complexity (funded by the Austrian Science Fund,
Grant DOI 10.55776/AR821), Lechner’s work is embedded in a broader investigation of
availability, opacity, and the roles of AI in musical instruments. The talk will present a series
of studies examining the modelling process itself, the use of AI to control large-scale models,
and the use of tools that allow for a fine-grained loss of control over instruments.

Robosonic – Sonification as acoustic feedback for human-robot interaction
Elias Naphausen

Industrial robots tend to work in silence, drawing attention only when errors occur or
maintenance is required.
In collaborative settings where humans and cobots share space without cages, state
information is conveyed
through lights, screens, or alarms—interfaces largely designed for monitoring fixed
workflows. As machinelearning
control broadens motion beyond predefined paths, robot behaviour can become
less predictable,
challenging situational awareness.
This work explores continuous sonification of robot telemetry—e.g., joint positions and
velocities, end-effector
speed, and interaction forces—as a complementary feedback channel for human–robot
interaction (HRI). By
mapping task-relevant signals to perceptually meaningful sound parameters, the system
exposes the robot’s
internal state in real time. We investigate whether characteristic rhythms emerge for
specific tasks, whether
deviations become audibly salient, and whether impending events (e.g., rising load, nearlimit
speeds) can be
anticipated “by ear.”
The contribution is a sonification framework and design heuristics for industrial
manipulators that aim to
improve transparency, coordination, and safety in close-proximity collaboration. Initial
studies hypothesise
that continuous acoustic feedback can support quicker anomaly detection and reduce visual
load during
shared tasks.

Audio Ghosts: Application of auditory-based illusions in sound installations
Piotr Majdak, Florian Pausch, Veronika Mayer, Bernhard Gál

For a long time, auditory-based illusions have been fascinating psychoacoustic research,
even though such illusions were usually investigated in isolated scientific studies. However,
illusions can be applied as a compositional tool in sound installations by artists and the link
between the psychoacoustics and artistic applications is not fully understood yet. The artsbased
research project Audio Ghosts investigates auditory-based illusions deliberately
applied in sound installations, addressing the interesting link between psychoacoustic
research and sound installations. As sound installations are usually concerned with the
artistic design and experience of acoustic environments, Audio Ghosts especially considers
spatial effects created by auditory illusions. In this talk, we will introduce the goals of Audio
Ghosts and provide an update on its current progress. We will describe various aspects of
sound installations, categorize auditory-based illusions, and showcase examples of
applications using auditory-based illusions in compositions and sound installations. Finally,
compositions, installations, and tools created within Audio Ghosts will be presented and
discussed.

Keynote: Musicality and Bioacoustics
Marisa Hoeschele

Where do music and language come from? Both appear to have their roots in biology, and
some aspects of both are shared with other species. Other species sing, dance, and
communicate information through the auditory domain. Here I will discuss what kind of
things we have learned about what makes audio such an important domain in our species
by providing examples of work we have done comparing humans to other animals. I will
show that 1) Both our drive towards rhythm and tonality in scales are likely driven by both
biology and the physics of sound that is shared with other species. 2) We also can recognize
emotional information from animals and they can recognize it from us because of how we
use our voices. 3) It may be the case that other animals have abilities that are closer to
language than we once thought. Overall, I hope that presenting this work will cause some
reflection about the deeper meaning of audio in both human and non-human life.

FlowSwing
Giorgios Sioros

Most music software organises rhythm around isochronous beats, treating timing as a
deviation from a fixed pulse. This creates a bias in audio environments that favours rigid
isochronous grids, obscuring the structural role and variability of the systematic nonisochronous
patterns found in many musical traditions and limiting how rhythmic
complexity can be represented and explored by producers and performers. Non-
Isochronous (NI) Grids offer an alternative framework that integrates dynamic nonisochronous
timing with structural organisation: they generate rhythmic structures through
the maximally even distribution of beats of two distinct durations.

FlowSwing implements NI Grids in Max/MSP as a tool for music production, sound design,
and composition. Audio and control signals can be time-warped continuously and aligned to
NI Grids, enabling performers and producers to smoothly move between isochronous and
non-isochronous organisations in real time. Beyond individual rhythm streams FlowSwing
models complex interactions between multiple NI Grids. This allows intricate relationships
and emergent rhythmic textures to be created, modulated and explored dynamically. By
embedding this theoretical framework in practice, FlowSwing treats rhythm itself as
malleable material shaped by structural models, opening up new possibilities in music and
sound production.

Sound as Score – Auditory modelling in the context of audio-score compositions
Katharina Pollack, Piotr Majdak, Barbara Maria Neu, Elisabeth Shimana

Audio scores are the auditory equivalent of the typically
visual music notation. Despite an a-priori knowledge about the various
parts of a musical piece, the interpretation of a musician presented
with audio scores heavily depends on their auditory system and their
auditory environment. In this talk, to demonstrate the application of
audio scores, we will present recordings from a concert of Elisabeth
Schimana’s “Virus #3.6” in St. Pölten’s cathedral performed by the Black
Page orchestra as well as isolated sound examples from clarinet
recordings in the lab. Further, we will present the general concept of
auditory modelling underlying the interaction between room, artists, and
sound. We will elaborate on the perception-action cycle using discrete
processing stages and cover current limits of said stages. Then, we will
discuss peripheral auditory models considered being appropriate
candidates to describe these perceptual effects, focusing on timing,
pitch, and spectral parameters. Finally, we relate isolated parts of
audio-score recordings to proposed auditory models and discuss the
conceptual incorporation of new sound examples into the proposed models.

"Non/Repeat: Three Case Studies of Non-linear Live-Music Practices"
Maria Kallionpää

The dramatic arc is a key factor on determining how to approach a musical composition
from the perspectives of a composer, performer, or music analyst. While most classical
music repertoires rely on established music forms such as symphony or sonata form,
composers have always experimented with alternative concepts of form, sometimes even
stripping them of any kind of predetermined structure. Eventually it is up to the composer
to define which parameters to set beforehand and which ones to leave up to chance or
depending on the decisions of the performers or the listeners. Non-linear music can be seen
as a response to such considerations: such works can range from musical games, generative,
aleatoric or algorithmic compositions up to total indeterminacy, possibly involving complete
or limited audience participation. Instead of yielding a lexicon on all the existing non-linear
music forms, the authors pursue to create knowledge on the ideas and techniques
surrounding them. We will discuss how non-linearity appears in musical compositions and
how the creative inputs of a live performance are distributed between the composer,
performers, and the listeners. This also raises a question on the authorship of the completed
work: how much of it was created by the composer, and what creative decisions stem from
the interpretations of the performer, or from the reactions of the audience members? We
will approach non-linear music from three different perspectives: through gamification,
audience participation, and audience reception.

Challenges in Mouse Ultrasonic Vocalization Analysis
Reyhaneh Abbasi

Mouse ultrasonic vocalizations (USVs) are widely used to study communication, behavior,
and neurological processes. Accurate analysis of these signals depends on robust tracking of
their frequency contours, or ridges, which encode information about vocalization
complexity. Existing tracking methods perform well on high-quality recordings but often
miss faint or low–signal-to-noise-ratio USV segments and fail to capture harmonics, limiting
their usefulness for automated analysis. In this talk, I will present a signal processing
approach developed to address these challenges, combining spectrogram and constant-Q
transform analysis with ridge tracking to improve contour detection and better capture
harmonic components, even in noisy recordings. Examples from experimental data will
illustrate ongoing challenges, such as tracking very faint or overlapping USVs in noisy
recordings, and highlight potential signal processing and machine learning strategies for
addressing them.

How to Filterbank: Audio Representations for Neural Networks
Daniel Haider and Felix Perfler

Should audio representations be learned or engineered? Addressing this question, three
paradigms are commonly used for representing raw audio in neural network architectures:
(i) fixed filterbanks such as the STFT, (ii) learned filterbanks implemented as 1D
convolutional layers, and (iii) hybrid approaches combining (i) and (ii). In this talk, we give a
brief overview of these three approaches, highlight advantages, and present solutions to
shortcomings, together with concrete implementations from our HybrA Python package.

Immersive Sound in Architecturally Complex Spaces
Fabio Kaiser

Die Umsetzung immersiver Audiosysteme in komplexen Räumen wie Theatern oder
Opernhäusern stellt besondere Herausforderungen an die Systemkonfiguration und
akustische Integration. Unterschiedliche Hörzonen, teilweise mit Fill-Systemen und variable
Anforderungen an Klangquellen erfordern flexible und automatisierbare Lösungen. Der
Vortrag beleuchtet zentrale Fragestellungen: Wie lassen sich einerseits bühnenbezogene
Quellen und andererseits Surround-Inhalte in einem gemeinsamen Lautsprechersystem
abbilden? Welche Strategien ermöglichen nahtlose Übergänge zwischen Front- und
Raumklang? Welche Ansätze sind erforderlich, um Delay- und Pegelstrukturen dynamisch an
Quelltyp und Hörposition anzupassen? Die Notwendigkeit, bühnenbezogene Quellen nicht
nur entlang der immersiven Front, sondern auf die echte Quellposition abzubilden macht
eine differenzierte Anwendung des Präzedenz-Effekts erforderlich. Ziel ist eine fundierte
Diskussion aktueller Konzepte in heterogenen Raumstrukturen.

Towards Modeling Elephant Hearing
Clara Hollomey, Jure Zeleznik, Daniel Haider, Peter Balazs, Angela Stöger, Matthias Zeppelzauer

Modeling species-dependent hearing is essential for understanding animal vocal
communication. Elephants primarily use rumbles—low-frequency, amplitude-modulated
vocalizations that travel long distances and convey information about identity, emotional
state, and social or reproductive context. Their amplitude modulation (AM) patterns are
believed to encode communicative intent.

To interpret these signals perceptually, we adapt human auditory models to the elephant
hearing range. We introduce an elephant-specific auditory filterbank as a foundation for
transferring psychoacoustic insights from humans to elephants. The model combines the
Greenwood frequency–place function with species-specific cochlear bandwidths, derived
analytically via differentiation of the center-frequency function. Frame theory ensures that
the filterbank preserves information across the time–frequency domain.

The resulting time–frequency representations reveal how rumbles and their AM structures
may be encoded within the elephant auditory system. These results will be presented in the
talk, along with perspectives on extending the framework toward roughness, sharpness, and
auditory compression measures, following the Lopez-Poveda cochlear model, to enable
perceptually meaningful comparisons between human and elephant hearing.