Data-driven microphone array shape identification in reverberant environments

With microphone arrays, we can obtain spatial information with the recorded signals, which can be used to ascertain sound source location and allow for source separation. A common assumption with the application of spatial audio algorithms is that these microphone arrays remain in place during the recording. However, there are many scenarios where this is not the case. Prior work with deformable arrays suggests beamforming improvements when knowledge of the array shape is known. While, solutions for determining the array shape have been formulated in anechoic environments, these solutions are not feasible in a reverberant setting, as time delay estimates are subject to error. This thesis proposes a data-driven solution for array shape identification in reverberant environments, using the relative transfer function as a fingerprint for the array configuration. Not only does this thesis justify the relative transfer function as a fingerprint in this manner, but it also details a semi-supervised model to learn array deformation parameters.

Name thesis:degree_name

M.S.

Level thesis:degree_level

Thesis

Discipline thesis:degree_discipline

Electrical & Computer Engr

Grantor

University of Illinois at Urbana-Champaign

Year dc:date

2023

Author dc:creator

• Sarkar, Kanad

Contributors dc:contributor

• Singer, Andrew C.

• Semi-Supervised Learning
• Spatial Audio
• Array Processing
• Deformable Arrays
• Manifold Learning

Statement dc:rights

• Copyright 2023 Kanad Sarkar

Language dc:language

en

Handle dc:identifier

https://hdl.handle.net/2142/122165

OAI identifier oai:identifier

oai:www.ideals.illinois.edu:2142/122165