Skip navigation
Skip navigation

Spatial dissection of a soundfield using spherical harmonic decomposition

Fahim, Abdullah

Description

A real-world soundfield is often contributed by multiple desired and undesired sound sources. The performance of many acoustic systems such as automatic speech recognition, audio surveillance, and teleconference relies on its ability to extract the desired sound components in such a mixed environment. The existing solutions to the above problem are constrained by various fundamental limitations and require to enforce different priors depending on the acoustic condition such as reverberation and...[Show more]

dc.contributor.authorFahim, Abdullah
dc.date.accessioned2020-04-30T23:11:58Z
dc.date.available2020-04-30T23:11:58Z
dc.identifier.otherb71498096
dc.identifier.urihttp://hdl.handle.net/1885/203506
dc.description.abstractA real-world soundfield is often contributed by multiple desired and undesired sound sources. The performance of many acoustic systems such as automatic speech recognition, audio surveillance, and teleconference relies on its ability to extract the desired sound components in such a mixed environment. The existing solutions to the above problem are constrained by various fundamental limitations and require to enforce different priors depending on the acoustic condition such as reverberation and spatial distribution of sound sources. With the growing emphasis and integration of audio applications in diverse technologies such as smart home and virtual reality appliances, it is imperative to advance the source separation technology in order to overcome the limitations of the traditional approaches. To that end, we exploit the harmonic decomposition model to dissect a mixed soundfield into its underlying desired and undesired components based on source and signal characteristics. By analysing the spatial projection of a soundfield, we achieve multiple outcomes such as (i) soundfield separation with respect to distinct source regions, (ii) source separation in a mixed soundfield using modal coherence model, and (iii) direction of arrival (DOA) estimation of multiple overlapping sound sources through pattern recognition of the modal coherence of a soundfield. We first employ an array of higher order microphones for soundfield separation in order to reduce hardware requirement and implementation complexity. Subsequently, we develop novel mathematical models for modal coherence of noisy and reverberant soundfields that facilitate convenient ways for estimating DOA and power spectral densities leading to robust source separation algorithms. The modal domain approach to the soundfield/source separation allows us to circumvent several practical limitations of the existing techniques and enhance the performance and robustness of the system. The proposed methods are presented with several practical applications and performance evaluations using simulated and real-life dataset.
dc.language.isoen_AU
dc.titleSpatial dissection of a soundfield using spherical harmonic decomposition
dc.typeThesis (PhD)
local.contributor.supervisorSamarasinghe, Prasanga
local.contributor.supervisorcontactu4801876@anu.edu.au
dc.date.issued2020
local.contributor.affiliationCollege of Engineering & Computer Science, The Australian National University
local.identifier.doi10.25911/5ec264755a3bc
local.identifier.proquestYes
local.identifier.researcherIDV-1986-2019
local.thesisANUonly.author7b5582ed-c59f-4f90-9e93-407365747052
local.thesisANUonly.title000000015620_TC_1
local.thesisANUonly.key4fd2f20b-c24e-db87-52b4-6387cb48b025
local.mintdoimint
CollectionsOpen Access Theses

Download

File Description SizeFormat Image
Fahim_PhD_Thesis.pdfThesis Material8.76 MBAdobe PDFThumbnail


Items in Open Research are protected by copyright, with all rights reserved, unless otherwise indicated.

Updated:  19 May 2020/ Responsible Officer:  University Librarian/ Page Contact:  Library Systems & Web Coordinator