Blind Source Separation: Advances in Theory, Algorithms and Applications

548 Pages·2014·19.565 MB·English

by  Ganesh R. Naik, Wenwu Wang (eds.)

Signals and Communication Technology
Ganesh R. Naik
Wenwu Wang Editors
Blind
Source
Separation
Advances in Theory, Algorithms and
Applications
Signals and Communication Technology
For furthervolumes:
http://www.springer.com/series/4748
Ganesh R. Naik Wenwu Wang
•
Editors
Blind Source Separation
Advances in Theory, Algorithms
and Applications
123
Editors
Ganesh R. Naik WenwuWang
Universityof Technology Universityof Surrey
Sydney Guildford
Australia UK
ISSN 1860-4862 ISSN 1860-4870 (electronic)
ISBN 978-3-642-55015-7 ISBN 978-3-642-55016-4 (eBook)
DOI 10.1007/978-3-642-55016-4
Springer Heidelberg NewYork Dordrecht London
LibraryofCongressControlNumber:2014940320
(cid:2)Springer-VerlagBerlinHeidelberg2014
Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpartof
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation,broadcasting,reproductiononmicrofilmsorinanyotherphysicalway,andtransmissionor
informationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar
methodology now known or hereafter developed. Exempted from this legal reservation are brief
excerpts in connection with reviews or scholarly analysis or material supplied specifically for the
purposeofbeingenteredandexecutedonacomputersystem,forexclusiveusebythepurchaserofthe
work. Duplication of this publication or parts thereof is permitted only under the provisions of
theCopyright Law of the Publisher’s location, in its current version, and permission for use must
always be obtained from Springer. Permissions for use may be obtained through RightsLink at the
CopyrightClearanceCenter.ViolationsareliabletoprosecutionundertherespectiveCopyrightLaw.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexempt
fromtherelevantprotectivelawsandregulationsandthereforefreeforgeneraluse.
While the advice and information in this book are believed to be true and accurate at the date of
publication,neithertheauthorsnortheeditorsnorthepublishercanacceptanylegalresponsibilityfor
anyerrorsoromissionsthatmaybemade.Thepublishermakesnowarranty,expressorimplied,with
respecttothematerialcontainedherein.
Printedonacid-freepaper
SpringerispartofSpringerScience+BusinessMedia(www.springer.com)
Preface
Blindsourceseparation(BSS)methodshavereceivedextensiveattentionoverthe
past two decades; thanks to its wide applicability in a number of areas such as
biomedical engineering, audio signal processing, and telecommunications. The
problem of source separation is an inductive inference problem, as only limited
information, e.g., the sensor observations, is available to infer the most probable
source estimates. The aim of BSS is to process these observations (acquired by
sensors or sensor arrays) in such a way that the original unknown source signals
areextractedby,e.g.,anadaptivesystem,orseparatedsimultaneouslyusing,e.g.,
a block (or batch)-based algorithm, without knowing or with very limited infor-
mation about the characteristics of the transmission channels through which the
sourcespropagatetothesensors.Independentcomponentanalysis(ICA)isoneof
theearlyandmostwidelyusedtechniquesforBSSbyrevealingthehiddenfactors
thatunderliethesetsofmeasurementsortheobservedsignals.Recently,anumber
of new techniques have been emerging in BSS, such as latent variable analysis,
non-negative matrix/tensor factorization (NMF/NTF), sparse component analysis,
dictionary learning, independent vector analysis, factor analysis, matrix comple-
tion, compressed sensing, empirical mode decomposition, and complex-valued
adaptivemethods.Atthesametime,theapplicationsofBSScontinuetogrowand
prosper in a number of areas, such as audio, speech, music, image, biomedical,
communication, and financial data analysis and processing.
This book aims to disseminate timely to the scientific community the new
developments in BSS spanning from theoretical frameworks, algorithmic devel-
opments,toavarietyofapplications. Thebookcoverssomeemerging techniques
in BSS, especially those developed recently, offering academic researchers and
practitionersacomprehensiveupdateaboutthenewdevelopmentinthisfield.The
book provides a forum for researchers to exchange their ideas, and to foster a
better understanding of the state of the art of the subject. We envisage that the
publication of this book will motivate new ideas and more cutting-edge research
activities in this area.
This book is intended for computer science and electronics engineers
(researchersandgraduatestudents)whowishtogetnovelresearchideasandsome
training in BSS, ICA, machine learning, artificial intelligence, and signal pro-
cessing applications. Furthermore, the research results previously scattered in
manyscientificarticlesworldwidearemethodicallycollectedandpresentedinthe
v
vi Preface
bookinaunifiedform.Asaresultofitstwofoldcharacter,thebookislikelytobe
of interest to researchers, engineers, and graduates who wish to learn the core
principles, methods, algorithms, and applications of BSS. Furthermore, the book
mayalsobeofbroaderinteresttoresearchersworkinginotherareasofscienceand
engineering, due to the multidisciplinary nature of this book.
The book is organized into two parts. Part I is devoted to recent developments
intheories,algorithms,andextensionsofBSS.Inthispart,wehavecollectednine
chapters with several novel contributions, namely, the idea of quantum ICA by
Yannick Deville and Alain Deville, the singularity-aware dictionary learning
approachforBSSbyXiaochenZhao,GuangyuZhou,WenwuWang,andWeiDai,
thetheoreticalresultsontheperformanceofcomplexICAbyBenediktLoeschand
Bin Yang, sub-band based BSS by Bo Peng and Wei Liu, independent vector
analysis for frequency domain BSS by Yanfeng Liang, Syed Mohsen Naqvi,
WenwuWang,andJonathonA.Chambers,sparsecomponentanalysisbyYannick
Deville,underdeterminedsourceseparationbyNikolaosMitianoudis,NMFbased
sourceseparationbyBinGaoandWaiLokWoo,andaBSSrelatedtopicofsource
localisation and tracking by Md Mashud Hyder and Kaushik Mahata. Part II
focuses on the various applications of BSS and its links to other relevant areas,
such as computational auditory scene analysis (CASA). We have gathered 10
chapters in this part. They are respectively, blind speech extraction algorithms by
HiroshiSaruwatariandRyoichiMiyazaki,combiningsuperdirectivebeamforming
andBSSforspeechseparationbyLinWang,HepingDing,andFuliangYin,ideal
ratio mask for CASA by Christopher Hummersone, Toby Stokes, and Tim
Brookes, monaural speech enhancement by Masoud Geravanchizadeh and Reza
Ahmadnia, background/foreground separation by Zafar Rafii, Antoine Liutkus,
and Bryan Pardo, NMF-based sparse coding for cochlear implants by Hongmei
Hu, Guoping Li, Mark E. Lutman, and Stefan Bleeck, brain signal analysis using
ICAbyRubenMartin-Clemente,BSSfortheanalysisoflarge-scaleomicdatasets
byAndrewE.Teschendorff,EmilieRenard,andPierreA.Absil,ICAforcomplex
domain source separation of communication signals by Ajay K. Kattepur and
Farook Sattar, and semi-blind source separation algorithms from non-invasive
electrophysiology to neuro-imaging by Camillo Porcaro and Franca Tecchio.
Wewouldliketothanktheauthorsfortheirexcellentsubmissions(chapters)to
this book, and their significant contributions to the review process, which have
helpedtoensurethehighqualityofthispublication.Withouttheircontributions,it
would have not been possible for the book to come successfully into existence.
January 2014 Ganesh R. Naik
Wenwu Wang
Contents
Part I Theory, Algorithms, and Extensions
1 Quantum-Source Independent Component Analysis
and Related Statistical Blind Qubit Uncoupling Methods. . . . . . . 3
Yannick Deville and Alain Deville
2 Blind Source Separation Based on Dictionary Learning:
A Singularity-Aware Approach. . . . . . . . . . . . . . . . . . . . . . . . . . 39
Xiaochen Zhao, Guangyu Zhou, Wei Dai and Wenwu Wang
3 Performance Study for Complex Independent
Component Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Benedikt Loesch and Bin Yang
4 Subband-Based Blind Source Separation
and Permutation Alignment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Bo Peng and Wei Liu
5 Frequency Domain Blind Source Separation Based
on Independent Vector Analysis with a Multivariate
Generalized Gaussian Source Prior. . . . . . . . . . . . . . . . . . . . . . . 131
Yanfeng Liang, Syed Mohsen Naqvi, Wenwu Wang
and Jonathon A. Chambers
6 Sparse Component Analysis: A General Framework
for Linear and Nonlinear Blind Source Separation
and Mixture Identification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151
Yannick Deville
7 Underdetermined Audio Source Separation Using
Laplacian Mixture Modelling . . . . . . . . . . . . . . . . . . . . . . . . . . . 197
Nikolaos Mitianoudis
vii
viii Contents
8 Itakura-Saito Nonnegative Matrix Two-Dimensional
Factorizations for Blind Single Channel Audio Separation. . . . . . 231
Bin Gao and Wai Lok Woo
9 Source Localization and Tracking: A Sparsity-Exploiting
Maximum a Posteriori Based Approach . . . . . . . . . . . . . . . . . . . 259
Md Mashud Hyder and Kaushik Mahata
Part II Applications
10 Statistical Analysis and Evaluation of Blind Speech
Extraction Algorithms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 291
Hiroshi Saruwatari and Ryoichi Miyazaki
11 Speech Separation and Extraction by Combining Superdirective
Beamforming and Blind Source Separation. . . . . . . . . . . . . . . . . 323
Lin Wang, Heping Ding and Fuliang Yin
12 On the Ideal Ratio Mask as the Goal of Computational
Auditory Scene Analysis. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 349
Christopher Hummersone, Toby Stokes and Tim Brookes
13 Monaural Speech Enhancement Based
on Multi-threshold Masking . . . . . . . . . . . . . . . . . . . . . . . . . . . . 369
Masoud Geravanchizadeh and Reza Ahmadnia
14 REPET for Background/Foreground Separation in Audio . . . . . . 395
Zafar Rafii, Antoine Liutkus and Bryan Pardo
15 Nonnegative Matrix Factorization Sparse Coding Strategy
for Cochlear Implants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413
Hongmei Hu, Guoping Li, Mark E. Lutman and Stefan Bleeck
16 Exploratory Analysis of Brain with ICA . . . . . . . . . . . . . . . . . . . 435
Rubén Martín-Clemente
17 Supervised Normalization of Large-Scale Omic Datasets
Using Blind Source Separation . . . . . . . . . . . . . . . . . . . . . . . . . . 465
Andrew E. Teschendorff, Emilie Renard and Pierre A. Absil
Contents ix
18 FebICA: Feedback Independent Component Analysis
for Complex Domain Source Separation
of Communication Signals . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 499
A. K. Kattepur and F. Sattar
19 Semi-blind Functional Source Separation Algorithm
from Non-invasive Electrophysiology to Neuroimaging. . . . . . . . . 521
Camillo Porcaro and Franca Tecchio
Part I
Theory, Algorithms, and Extensions