Private audio delivery in reverberant spaces

Liu, Yu-Jeh

Permalink

https://hdl.handle.net/2142/108006 Copy

Description

Title

Private audio delivery in reverberant spaces

Author(s)

Liu, Yu-Jeh

Issue Date

2020-05-11

Director of Research (if dissertation) or Advisor (if thesis)

Dokmanic, Ivan

Department of Study

Electrical & Computer Eng

Discipline

Electrical & Computer Engr

Degree Granting Institution

University of Illinois at Urbana-Champaign

Degree Name

M.S.

Degree Level

Thesis

Keyword(s)

• sound zone
• reverberation
• sound focusing
• private audio
• least-squares
• null space
• inverse problem
• soundfield
• room acoustics
• eavesdropper
• loudspeaker array

Abstract

We study the problem of delivering private audio to multiple users in a reverberant room. New methods are proposed which work with specially designed noise signals and make constructive use of reverberation. First, traditional sound zones problem formulation is introduced. In this formulation, a loudspeaker array is used to create isolated soundfields as designated by the user. Traditional approaches such as acoustic contrast control (ACC) and pressure matching (PM) were devised as solutions to the sound zones problem, and they were proven to be effective. One aspect that has not been addressed by these methods is the privacy/security issue. With eavesdroppers present in the room, traditional approaches to sound zones problems do not ensure secure delivery of audio that prevents the eavesdroppers from comprehending the contents. Next, new methods are thus introduced which address the added privacy requirement. Instead of considering reverberation as unwelcome, the proposed formulation takes advantage of the multi-path nature and leverages the random-like echoes to simultaneously achieve sound focusing and eavesdropping prevention. Two methods are introduced. The first one is based on direct least-squares optimization; we refer to it as the Least-Squares (LS) method. The second one, adopted from the wireless communication literature, exploits the null space of the multiple-input-multiple-output channel matrix; we refer to it as the Null Space (NS) method. The two methods are evaluated and compared in numerical and real-world experiments. Both methods are shown to achieve sound focusing as well as very low signal-to-noise ratio (SNR) outside the focusing spots. The NS method provides sharper intelligibility drop which results in smaller, more spatially refined sound focusing spots. Subject to the usual limitations of sound zone methods such as long computation time or the requirement to know the impulse responses, both methods are proven to provide new solutions to sound zone problems with privacy constraints.

Graduation Semester

2020-05

Type of Resource

Thesis

Permalink

http://hdl.handle.net/2142/108006

Copyright and License Information

Copyright 2020 Yu-Jeh Liu

Owning Collections