Independent Component Analysis and Blind Signal Separation

In many situations found both in Nature and in human-built systems, a set of mixed signalsisobserved(frequentlyalsowithnoise),anditisofgreatscienti?candtech- logicalrelevanceto beableto isolateor separatethemso thattheinformationin each ofthesignalscanbeutilized. Blind source separation (BSS) research is one of the more interesting emerging ?eldsnowadaysinthe?eldofsignalprocessing.Itdealswiththealgorithmsthatallow therecoveryoftheoriginalsourcesfromasetofmixturesonly.Theadjective blind is applied becausethe purposeis to estimate the originalsourceswithoutany a priori knowledgeabouteitherthesourcesorthemixingsystem.Mostofthemodelsemployed in BSS assume the hypothesisabout the independenceof the original sources. Under this hypothesis,a BSS problemcan be consideredas a particularcase of independent componentanalysis(ICA),alineartransformationtechniquethat,startingfromam- tivariate representation of the data, minimizes the statistical dependence between the componentsoftherepresentation.Itcan beclaimed thatmostoftheadvancesin ICA havebeenmotivatedbythesearchforsolutionstotheBSSproblemand,theotherway around,advancesinICAhavebeenimmediatelyappliedtoBSS. ICA and BSS algorithms start from a mixture model, whose parameters are - timated from the observed mixtures. Separation is achieved by applying the inverse mixturemodelto theobservedsignals(separatingorunmixingmodel).Mixturem- els usually fall into three broad categories: instantaneous linear models, convolutive modelsandnonlinearmodels,the?rstonebeingthesimplestbut,in general,notnear realisticapplications.Thedevelopmentandtestofthealgorithmscanbeaccomplished throughsyntheticdataorwithreal-worlddata.Obviously,themostimportantaim(and mostdif?cult)istheseparationofreal-worldmixtures.BSSandICAhavestrongre- tionsalso,apartfromsignalprocessing,withother?eldssuchasstatisticsandarti?cial neuralnetworks. As long as we can ?nd a system that emits signals propagated through a mean, andthosesignalsarereceivedbyasetofsensorsandthereisaninterestinrecovering the originalsources,we have a potential?eld ofapplication forBSS and ICA. Inside thatwiderangeofapplicationswecan?nd,forinstance:noisereductionapplications, biomedicalapplications,audiosystems,telecommunications,andmanyothers. This volume comes out just 20 years after the ?rst contributionsin ICA and BSS 1 appeared . Thereinafter,the numberof research groupsworking in ICA and BSS has been constantly growing, so that nowadays we can estimate that far more than 100 groupsareresearchinginthese?elds. Asproofoftherecognitionamongthescienti?ccommunityofICAandBSSdev- opmentstherehavebeennumerousspecialsessionsandspecialissuesinseveralwell- 1 J.Herault, B.Ans, Circuits neuronaux à synapses modi?ables: décodage de messages c- posites para apprentissage non supervise , C.R. de l Académie des Sciences, vol. 299, no. III-13,pp.525 528,1984.

Includes supplementary material: sn.pub/extras

Klappentext

This book constitutes the refereed proceedings of the 5th International Conference on Independent Component Analysis and Blind Source Separation, ICA 2004, held in Granada, Spain, in September 2004.

The 156 revised papers presented were carefully reviewed and selected from 203 submissions. The papers are organized in topical sections on theory and foundations, linear models, covolutive models, nonlinear models, speech processing applications, image processing applications, biomedical applications, and other applications.

Inhalt
Theory and Fundamentals.- A FastICA Algorithm for Non-negative Independent Component Analysis.- Blind Source Separation by Adaptive Estimation of Score Function Difference.- Exploiting Spatiotemporal Information for Blind Atrial Activity Extraction in Atrial Arrhythmias.- Gaussianizing Transformations for ICA.- New Eigensystem-Based Method for Blind Source Separation.- Optimization Issues in Noisy Gaussian ICA.- Optimization Using Fourier Expansion over a Geodesic for Non-negative ICA.- The Minimum Support Criterion for Blind Signal Extraction: A Limiting Case of the Strengthened Young's Inequality.- Accurate, Fast and Stable Denoising Source Separation Algorithms.- An Overview of BSS Techniques Based on Order Statistics: Formulation and Implementation Issues.- Analytical Solution of the Blind Source Separation Problem Using Derivatives.- Approximate Joint Diagonalization Using a Natural Gradient Approach.- BSS, Classification and Pixel Demixing.- Blind Identification of Complex Under-Determined Mixtures.- Blind Separation of Heavy-Tailed Signals Using Normalized Statistics.- Blind Source Separation of Linear Mixtures with Singular Matrices.- Closely Arranged Directional Microphone for Source Separation.- Estimating Functions for Blind Separation when Sources Have Variance-Dependencies.- Framework of Constrained Matrix Gradient Flows.- Identifiability, Subspace Selection and Noisy ICA.- Improving GRNNs in CAD Systems.- Fisher Information in Source Separation Problems.- Localization of P300 Sources in Schizophrenia Patients Using Constrained BSS.- On the Estimation of the Mixing Matrix for Underdetermined Blind Source Separation in an Arbitrary Number of Dimensions.- On the Minimum ?1-Norm Signal Recovery in Underdetermined Source Separation.- On the Strong Uniqueness of Highly Sparse Representations from Redundant Dictionaries.- Reliability of ICA Estimates with Mutual Information.- Robust ICA for Super-Gaussian Sources.- Robustness of Prewhitening Against Heavy-Tailed Sources.- Simultaneous Extraction of Signal Using Algorithms Based on the Nonstationarity.- Space-Time Variant Blind Source Separation with Additive Noise.- The Use of ICA in Speckle Noise.- Theoretical Method for Solving BSS-ICA Using SVM.- Wavelet De-noising for Blind Source Separation in Noisy Mixtures.- Linear Mixture Models.- A Gaussian Mixture Based Maximization of Mutual Information for Supervised Feature Extraction.- Blind Separation of Nonstationary Sources by Spectral Decorrelation.- Delayed AMUSE A Tool for Blind Source Separation and Denoising.- Dimensionality Reduction in ICA and Rank-(R 1,R 2,...,R N ) Reduction in Multilinear Algebra.- Linear Multilayer Independent Component Analysis Using Stochastic Gradient Algorithm.- Minimax Mutual Information Approach for ICA of Complex-Valued Linear Mixtures.- Signal Reconstruction in Sensor Arrays Using Temporal-Spatial Sparsity Regularization.- Underdetermined Source Separation with Structured Source Priors.- A Grassmann-Rayleigh Quotient Iteration for Dimensionality Reduction in ICA.- An Approach of Moment-Based Algorithm for Noisy ICA Models.- Geometrical ICA-Based Method for Blind Separation of Super-Gaussian Signals.- A Novel Method to Recover N Sources from N-1 Observations and Its Application to Digital Communications.- A Sufficient Condition for Separation of Deterministic Signals Based on Spatial Time-Frequency Representations.- Adaptive Robust Super-exponential Algorithms for Deflationary Blind Equalization of Instantaneous Mixtures.- Application of Gaussian Mixture Models for Blind Separation of Independent Sources.- Asymptotically Optimal Blind Separation of Parametric Gaussian Sources.- Bayesian Approach for Blind Separation of Underdetermined Mixtures of Sparse Sources.- Blind Source Separation Using the Block-Coordinate Relative Newton Method.- Hybridizing Genetic Algorithms with ICA in Higher Dimension.- ICA Using Kernel Entropy Estimation with NlogN Complexity.- Soft-LOST: EM on a Mixture of Oriented Lines.- Some Gradient Based Joint Diagonalization Methods for ICA.- Underdetermined Independent Component Analysis by Data Generation.- Convolutive Models.- Batch Mutually Referenced Separation Algorithm for MIMO Convolutive Mixtures.- Frequency Domain Blind Source Separation for Many Speech Signals.- ICA Model Applied to Multichannel Non-destructive Evaluation by Impact-Echo.- Monaural Source Separation Using Spectral Cues.- Multichannel Speech Separation Using Adaptive Parameterization of…