Colloquium - Details

You will receive information about presentations in time if you subscribe to the newsletter of the Colloquium Communications Technology.

All interested students are cordially invited, registration is not required.

Promotionsvortrag: Signal Processing Concepts for User Movement in Scene-Based Spatial Audio

Maximilian Kentgens
Donnerstag, 29. Juni 2023
11:30 Uhr

This dissertation addresses prospective immersive communication, telepresence, and multimedia systems, in which a user moves around virtually in a remote or recorded acoustic scene. More specifically, the problem of sound field translation of a single-perspective higher-order Ambisonics (HOA) acoustic scene representation is considered.

Recent standardization activities and ongoing research have considered two competing approaches for immersive spatial audio formats: object-based audio on the one hand and scene-based audio on the other hand, each of which has different advantages and disadvantages. Object-based audio is the preferred choice for artist- or computer-generated acoustic scenes. In contrast, scene-based audio is better suited to capture an actual acoustic scene in its entirety, which makes it well suited in the above-mentioned applications. Scene-based audio is commonly realized using HOA, i.e., the spherical-harmonics-based (SH) representation of the sound field. This approach allows the direct acquisition of real-world content using spherical microphone arrays. In contrast to object-based audio that inherently supports user interaction in six degrees of freedom (6DoF), HOA recordings allow for user rotation only (3DoF). HOA lacks a built-in possibility to adapt the listener’s position in space. This dissertation aims to alleviate this shortcoming of scene-based audio by novel HOA signal processing methods exploiting signal characteristics and properties of the human auditory system.

State-of-the-art methods for HOA sound field translation in literature are either limited to very small displacements or laboratory conditions with almost no reverberation. Others require additional microphones or are limited to first-order Ambisonics. In contrast, the present dissertation considers the application of user movement in immersive audio with HOA signal content recorded at a single position in space under adverse conditions with reverberation. Methodologically, the problem is approached via a theoretical elaboration of signal processing concepts based on acoustic considerations and psychoacoustic evidence. The main focus is on the mathematical and simulative analysis and the comparison of the proposed approaches. The work is rounded off by a perceptual study showing the potential of the methods under realistic boundary conditions.