[ 7. Januar 2024 ]

DEGEM News – FWD – [ak-discourse] Einladung zum Forschungskolloquium am 9.1.2024

Von: Fabian Brinkmann via al discourse
Datum: Thu, 4 Jan 2024
Betreff: [ak-discourse] Einladung zum Forschungskolloquium am 9.1.2024

Liebe Freund*innen der Audiokommunikation,

am kommenden Dienstag, 9.1.2024, 16:00 Uhr im Raum EN 324 präsentiert Daniel Wujecki seine Masterarbeit zum Thema

„Differentiable acoustic path-tracing for the optimization of room geometry“.

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/65603959784?pwd=ZEs1SkVrbFJzZ0pXeG9TSGp3YkI3UT09
Meeting-ID: 656 0395 9784
Kenncode: 20231009

Mit herzlichen Grüßen und fröhliche Weihnachten
Fabian Brinkmann & Stefan Weinzierl

Differentiable rendering is used for the optimization of scene parameters in computer graphics. This thesis transfers these methods to acoustics, allowing the optimization of scene parameters, such as material properties and geometry, to achieve specific acoustic characteristics in room environments like concert halls and recording studios. While some wave phenomena of sound are neglected, it is necessary to track the time for simulation of sound propagation. Differentiable rendering for light transport relies on two major methods, that need to be adapted for the propagation of sound. First, the method of Path Replay Backpropagation is transferred to efficiently compute time-sensitive scene parameter gradients and avoid backpropagation of extensive computation graphs. Second, to handle discontinuities caused by some scene parameters, such as geometry, the method of Unbiased Warped-Area sampling is aligned with the needs of acoustic path tracing. Since the temporal dimension of sound poses a additional discontinuity, Unbiased Warped-Area Sampling is complemented with a additional Gaussian convolution along the temporal dimension. In a comprehensive set of experiments the capabilities of differentiable acoustic path tracing are thoroughly tested. Real-world applications further validate the effectiveness of the proposed methods.


Fabian Brinkmann
Form of address: Fabian (no pronouns)

Senior Researcher
Audio Communication Group
Technische Universität Berlin

mail: fabian.brinkmann@tu-berlin.de
phon: +49 30 314 70336