[ 15. Juni 2024 ]

DEGEM News – FWD – [ak-discourse] Einladung zum Forschungskolloquium am 18.06.2024

Von: Andreev, Kirill via ak discourse
Datum: Fri, 14 Jun 2024
Betreff: [ak-discourse] Einladung zum Forschungskolloquium am 18.06.2024

Liebe Freunde der Audiokommunikation,

am kommenden Dienstag, den 18.06.2024, um 16:00 Uhr im Raum EN 324 präsentiert Zeyu Yang seine Masterarbeit zum Thema

„Autogenous Spatialization for Arbitrary Loudspeaker Setups“.

Im Anschluss präsentiert Or Berebi, Austausch-PhD-Student aus der Ben Gurion Universität Negev, seinen Vortrag zum Thema

„Perceptual optimization of low order Spherical Harmonics Representations using Neural Networks“.

Dazu möchten wir Sie herzlich einladen. Eine Kurzzusammenfassung finden Sie am Ende dieser E-Mail.

Der Zoom-Link für auswärtige Gäste lautet:

https://tu-berlin.zoom.us/j/64026853839?pwd=VzZWUU1ldjZlem9ZdlUyNUNFaHZxUT09

Meeting-ID: 640 2685 3839

Kenncode: 704269

Mit herzlichen Grüßen

Kirill Andreev & Stefan Weinzierl

Zeyu Yang: Autogenous Spatialization for Arbitrary Loudspeaker Setups

This thesis introduces Zerr*, a novel approach to spatial music production that navigates between conventional sound spatialization and spatial sound synthesis. Following an extensive review of sound spatialization techniques, as well as historical and contemporary developments in spatial music, a tailored coordinate system has been developed to categorize Zerr* alongside existing spatial music authoring tools. Zerr* employs an innovative algorithmic framework that leverages the intrinsic properties of the audio signal, coupled with a special mapping system, to autonomously distribute audio to arbitrary loudspeaker setups. This approach facilitates dynamic, context-sensitive spatialization and enables unique spatial sound synthesis effects. This approach effectively circumvents the limitations imposed by traditional spatialization techniques, thus introducing new sonic experiences and creative paradigms. The core modules of Zerr* are implemented in C++ and are further extended as a Pure Data package and JACK clients. The system is designed to integrate seamlessly into existing creative ecosystems, supporting real-time audio manipulation and sample-level spatialization. This functionality has proven particularly effective for live performances and improvisational contexts. Comprehensive listening tests conducted with participants from diverse backgrounds have validated the system’s effectiveness. Their feedback has confirmed the system’s innovative approach and its practical applicability in real-world settings.

Or Berebi: Perceptual optimization of low order Spherical Harmonics Representations using Neural Networks

Head-related transfer functions (HRTF) describe the direction-dependent sound arriving at the ears of a listener and are needed for 3D audio reproduction over headphones. In the context of headphone-based spatial audio, the Ambisonics format is widely employed, requiring HRTF data to be often represented in a low-order spherical harmonics order to match signals from an order restrcted microphone array. This research presents a learning-based approach aimed at minimizing perceptually motivated error terms associated with the low-order representation of HRTFs. The proposed method considers error measures covering various aspects of spatial hearing, including coloration, localization, and diffuse-field coherence. These factors are combined in a multi-object optimization by weighting them across space and frequency. This aggregated error serves as a loss function for a neural network, which takes an initial low-order representation of the HRTF and transforms it into the desired representation. Preliminary findings indicate that minimizing this combined error leads to enhancements compared to the widely used MagLS preprocessing method in Ambisonics.

Kirill Andreev

__

Sekretariat
Office

Technische Universität Berlin
Fakultät I / Inst. f. Sprache u. Kommunikation
FG Audiokommunikation / Sekr. EN 8 Raum 321

Faculty I / Institute for Language & Communication

Audio Communication Group / Office No. EN 8 Room 321

Einsteinufer 17, 10587 Berlin
GERMANY

+49 30 314-22236

kirill.andreev@tu-berlin.de
https://www.tu.berlin/ak