Novel techniques for improving the performance of personal sound zone systems
Novel techniques for improving the performance of personal sound zone systems
Personal sound zones (PSZ) systems aim to reproduce, through a set of loudspeakers, dedicated audio contents in specific zones, called bright zones, while minimising acoustic energy leakages into other zones within the same environment, called dark zones. This is achieved by filtering loudspeaker signals to create constructive and destructive interference in different regions of the space. As the first contribution of this thesis, the multi-objective optimisation problem (MOOP) framework is introduced and some solution strategies are proposed. Moreover, the acoustic contrast control (ACC) and pressure matching (PM) formulations of the PSZ problem are shown to represent particular instances of a MOOP. Additionally, it is established that for these two main formulations, more advanced MOOP solution strategies are unnecessary to explore the entire optimal solutions front. As the second contribution of this thesis, the effects of several pressure matching target sound field on the PSZ performance and robustness are compared. The first proposed target field prevents the PSZ system from focusing its effort on unessential tasks such as the dereverberation of the audio signal. The second, more advanced target field is obtained by optimising the magnitude and phase relation (ILD and IPD) of target sound pressure signals at different control points. Results obtained with real measure- ments also indicate a substantial reduction in array effort when employing optimised targets, except in the extreme regions of the audio spectrum. Results of simulations and experiments obtained with one compact loudspeaker array in the interior of a real car are presented, as well as a study on the robustness to perturbations of the solutions obtained with the various choices of target acoustic fields. To improve the robustness of PSZ systems to any environmental perturbation, regularisation techniques have been applied to the optimal filters, at the known cost of reduced acoustic contrast between the various zones. To address this challenge, the third contribution of this thesis consists of a novel theoretical framework which aims to combine the measurements made for various states of the PSZ system, reflecting variations in environmental conditions, in order to design a single set of optimal filters to be applied to the loudspeaker driving signals. This framework is then applied to both the PM and ACC techniques, leading to the introduction of two novel soundfield control techniques: the Multi-States Pressure Matching (MS- PM) and the Multi-States Acoustic Contrast Control (MS-ACC). Results of experiments obtained with two compact loudspeaker arrays in the interior of a real car are presented. The robustness performances of the PSZ system obtained with both MS-PM and MS-ACC techniques are compared with those obtained with the two conventional PM and ACC ones. As the fourth contribution of this thesis, a novel strategy for dynamic and spectral optimisation of audio content in any PSZ system is introduced. The primary goal of this process is to ensure a satisfactory perceived isolation level in both zones, regardless of the set of filters obtained using PM or ACC formulations. Experimental investigations conducted in a car cabin, utilising two 8-loudspeaker arrays and two binaural heads, have demonstrated the effectiveness of this optimisation process in redistributing target- to-interferer ratio (TIR) or extended short-time objective intelligibility (ESTOI) between the zones
University of Southampton
Gallian, Wilfried
d5111637-9d8b-42b0-b014-f5bc9ef342e6
February 2024
Gallian, Wilfried
d5111637-9d8b-42b0-b014-f5bc9ef342e6
Fazi, Filippo
e5aefc08-ab45-47c1-ad69-c3f12d07d807
Gallian, Wilfried
(2024)
Novel techniques for improving the performance of personal sound zone systems.
Doctoral Thesis, 146pp.
Record type:
Thesis
(Doctoral)
Abstract
Personal sound zones (PSZ) systems aim to reproduce, through a set of loudspeakers, dedicated audio contents in specific zones, called bright zones, while minimising acoustic energy leakages into other zones within the same environment, called dark zones. This is achieved by filtering loudspeaker signals to create constructive and destructive interference in different regions of the space. As the first contribution of this thesis, the multi-objective optimisation problem (MOOP) framework is introduced and some solution strategies are proposed. Moreover, the acoustic contrast control (ACC) and pressure matching (PM) formulations of the PSZ problem are shown to represent particular instances of a MOOP. Additionally, it is established that for these two main formulations, more advanced MOOP solution strategies are unnecessary to explore the entire optimal solutions front. As the second contribution of this thesis, the effects of several pressure matching target sound field on the PSZ performance and robustness are compared. The first proposed target field prevents the PSZ system from focusing its effort on unessential tasks such as the dereverberation of the audio signal. The second, more advanced target field is obtained by optimising the magnitude and phase relation (ILD and IPD) of target sound pressure signals at different control points. Results obtained with real measure- ments also indicate a substantial reduction in array effort when employing optimised targets, except in the extreme regions of the audio spectrum. Results of simulations and experiments obtained with one compact loudspeaker array in the interior of a real car are presented, as well as a study on the robustness to perturbations of the solutions obtained with the various choices of target acoustic fields. To improve the robustness of PSZ systems to any environmental perturbation, regularisation techniques have been applied to the optimal filters, at the known cost of reduced acoustic contrast between the various zones. To address this challenge, the third contribution of this thesis consists of a novel theoretical framework which aims to combine the measurements made for various states of the PSZ system, reflecting variations in environmental conditions, in order to design a single set of optimal filters to be applied to the loudspeaker driving signals. This framework is then applied to both the PM and ACC techniques, leading to the introduction of two novel soundfield control techniques: the Multi-States Pressure Matching (MS- PM) and the Multi-States Acoustic Contrast Control (MS-ACC). Results of experiments obtained with two compact loudspeaker arrays in the interior of a real car are presented. The robustness performances of the PSZ system obtained with both MS-PM and MS-ACC techniques are compared with those obtained with the two conventional PM and ACC ones. As the fourth contribution of this thesis, a novel strategy for dynamic and spectral optimisation of audio content in any PSZ system is introduced. The primary goal of this process is to ensure a satisfactory perceived isolation level in both zones, regardless of the set of filters obtained using PM or ACC formulations. Experimental investigations conducted in a car cabin, utilising two 8-loudspeaker arrays and two binaural heads, have demonstrated the effectiveness of this optimisation process in redistributing target- to-interferer ratio (TIR) or extended short-time objective intelligibility (ESTOI) between the zones
Text
Wilfried_s_PHD_Thesis_1_PDF/A
- Version of Record
Restricted to Repository staff only until 2 November 2024.
Text
Final-thesis-submission-Examination-Mr-Wilfried-Gallian (1)
Restricted to Repository staff only
More information
Published date: February 2024
Identifiers
Local EPrints ID: 486791
URI: http://eprints.soton.ac.uk/id/eprint/486791
PURE UUID: 78abbc87-4f22-4c1d-a94c-c437ac431c49
Catalogue record
Date deposited: 06 Feb 2024 17:39
Last modified: 17 Apr 2024 01:41
Export record
Download statistics
Downloads from ePrints over the past year. Other digital versions may also be available to download e.g. from the publisher's website.
View more statistics