Showing papers on "Independent component analysis published in 2014"
TL;DR: FIX ("FMRIB's ICA-based X-noiseifier"), which provides an automatic solution for denoising fMRI data via accurate classification of ICA components, and is being used in the default rfMRI processing pipeline for generating HCP connectomes.
1,565 citations
TL;DR: This overview article presents ICA, and then its generalization to multiple data sets, IVA, both using mutual information rate, and presents conditions for the identifiability of the given linear mixing model and derive the performance bounds.
Abstract: Starting with a simple generative model and the assumption of statistical independence of the underlying components, independent component analysis (ICA) decomposes a given set of observations by making use of the diversity in the data, typically in terms of statistical properties of the signal. Most of the ICA algorithms introduced to date have considered one of the two types of diversity: non-Gaussianity?i.e., higher-order statistics (HOS)?or, sample dependence. A recent generalization of ICA, independent vector analysis (IVA), generalizes ICA to multiple data sets and adds the use of one more diversity, dependence across multiple data sets for achieving an independent decomposition, jointly across multiple data sets. Finally, both ICA and IVA, when implemented in the complex domain, enjoy the addition of yet another type of diversity, noncircularity of the sources?underlying components. Mutual information rate provides a unifying framework such that all these statistical properties?types of diversity?can be jointly taken into account for achieving the independent decomposition. Most of the ICA methods developed to date can be cast as special cases under this umbrella, as well as the more recently developed IVA methods. In addition, this formulation allows us to make use of maximum likelihood theory to study large sample properties of the estimator, derive the Cram?r?Rao lower bound (CRLB) and determine the conditions for the identifiability of the ICA and IVA models. In this overview article, we first present ICA, and then its generalization to multiple data sets, IVA, both using mutual information rate, present conditions for the identifiability of the given linear mixing model and derive the performance bounds. We address how various methods fall under this umbrella and give examples of performance for a few sample algorithms compared with the performance bound. We then discuss the importance of approaching the performance bound depending on the goal, and use medical image analysis as the motivating example.
184 citations
TL;DR: The experimental results show that the proposed KDICA compares favorably with existing methods, and a new method of fault diagnosis, a non-linear contribution plot is developed for KDICA.
130 citations
01 Jan 2014
TL;DR: This paper examines the audio source separation problem using the general framework of Independent Component Analysis, and explores the case that the auditory scene is modeled as instantaneous mixtures of the auditory objects, to establish the basic tools for the analysis.
Abstract: In this paper we examine the audio source separation problem using the general framework of Independent Component Analysis (ICA). For the greatest part of the analysis, it has been assumed that equal number of sensors and sound objects. Firstly, it explores the case that the auditory scene is modeled as instantaneous mixtures of the auditory objects, to establish the basic tools for the analysis. The case of real room recordings, modeled as convolutive mixtures of the auditory objects, is then introduced. A novel Fast ICA framework is introduced, using two possible implementations. A great number of audio source separation problems can be addressed successfully using Independent Component Analysis. And concludes by highlighting some of the as yet unsolved problems to tackle the actual audio source separation problem in full.
122 citations
TL;DR: Wang et al. as discussed by the authors proposed a novel output-only damage identification method based on the unsupervised blind source separation (BSS) technique termed independent component analysis (ICA), which is discovered that ICA biases to extract sparse component, which typically indicates damage, from the observed mixture signals.
120 citations
TL;DR: An integration of principal components analysis (PCA) and independent component analysis (ICA) on transient thermal videos has been proposed, which enables spatial and temporal patterns to be extracted according to the transient response behavior without any training knowledge.
Abstract: Eddy current pulsed thermography (ECPT) is implemented for detection and separation of impact damage and resulting damages in carbon fiber reinforced plastic (CFRP) samples. Complexity and nonhomogeneity of fiber texture as well as multiple defects limit detection identification and characterization from transient images of the ECPT. In this paper, an integration of principal component analysis (PCA) and independent component analysis (ICA) on transient thermal videos has been proposed. This method enables spatial and temporal patterns to be extracted according to the transient response behavior without any training knowledge. In the first step, using the PCA, the data is transformed to orthogonal principal component subspace and the dimension is reduced. Multichannel morphological component analysis, as an ICA method, is then implemented to deal with the sparse and independence property for detecting and separating the influences of different layers, defects, and their combination information in the CFRP. Because different transient behaviors exist, multiple types of defects can be identified and separated by calculating the cross-correlation of the estimated mixing vectors between impact the ECPT sequences and nondefect ECPT sequences.
110 citations
TL;DR: A two‐stage anomaly detection strategy based on multiple distributed sensors located throughout the network, based on a binary classification scheme (detection is casted into an anomalous/normal classification problem) driven by machine learning‐inferred decision trees.
Abstract: Network anomalies, circumstances in which the network behavior deviates from its normal operational baseline, can be due to various factors such as network overload conditions, malicious/hostile activities, denial of service attacks, and network intrusions. New detection schemes based on machine learning principles are therefore desirable as they can learn the nature of normal traffic behavior and autonomously adapt to variations in the structure of 'normality' as well as recognize the significant deviations as suspicious or anomalous events. The main advantages of these techniques are that, in principle, they are not restricted to any specific environment and that they can provide a way of detecting unknown attacks. Detection performance is directly correlated with the traffic model quality, in terms of ability of representing the traffic behavior from its most characterizing internal dynamics. Starting from these ideas, we developed a two-stage anomaly detection strategy based on multiple distributed sensors located throughout the network. By using Independent Component Analysis, the first step, modeled as a Blind Source Separation problem, extracts the fundamental traffic components the 'source' signals, corresponding to the independent traffic dynamics, from the multidimensional time series incoming from the sensors, corresponding to the perceived 'mixed/aggregate' effect of traffic on their interfaces. These components will be used to build the baseline traffic profiles needed in the second supervised phase, based on a binary classification scheme detection is casted into an anomalous/normal classification problem driven by machine learning-inferred decision trees. Copyright © 2013 John Wiley & Sons, Ltd.
89 citations
Proceedings Article•
29 May 2014TL;DR: In this paper, it was shown that a mixture with known identical covariance matrices whose number of components is a polynomial of any fixed degree in the dimension n is polynomially learnable as long as a certain non-degeneracy condition on the means is satisfied.
Abstract: In this paper we show that very large mixtures of Gaussians are efficiently learnable in high dimension. More precisely, we prove that a mixture with known identical covariance matrices whose number of components is a polynomial of any fixed degree in the dimension n is polynomially learnable as long as a certain non-degeneracy condition on the means is satisfied. It turns out that this condition is generic in the sense of smoothed complexity, as soon as the dimensionality of the space is high enough. Moreover, we prove that no such condition can possibly exist in low dimension and the problem of learning the parameters is generically hard. In contrast, much of the existing work on Gaussian Mixtures relies on low-dimensional projections and thus hits an artificial barrier. Our main result on mixture recovery relies on a new “Poissonization”-based technique, which transforms a mixture of Gaussians to a linear map of a product distribution. The problem of learning this map can be efficiently solved using some recent results on tensor decompositions and Independent Component Analysis (ICA), thus giving an algorithm for recovering the mixture. In addition, we combine our low-dimensional hardness results for Gaussian mixtures with Poissonization to show how to embed difficult instances of low-dimensional Gaussian mixtures into the ICA setting, thus establishing exponential information-theoretic lower bounds for underdetermined ICA in low dimension. To the best of our knowledge, this is the first such result in the literature. In addition to contributing to the problem of Gaussian mixture learning, we believe that this work is among the first steps toward better understanding the rare phenomenon of the “blessing of dimensionality” in the computational aspects of statistical inference.
79 citations
TL;DR: A new method to identify compound faults from measured mixed-signals from ensemble empirical mode decomposition (EEMD) method and independent component analysis (ICA) technique is proposed, which makes the fault features more easily extracted and more clearly identified.
Abstract: A Compound fault signal usually contains multiple characteristic signals and strong confusion noise, which makes it difficult to separate week fault signals from them through conventional ways, such as FFT-based envelope detection, wavelet transform or empirical mode decomposition individually. In order to improve the compound faults diagnose of rolling bearings via signals’ separation, the present paper proposes a new method to identify compound faults from measured mixed-signals, which is based on ensemble empirical mode decomposition (EEMD) method and independent component analysis (ICA) technique. With the approach, a vibration signal is firstly decomposed into intrinsic mode functions (IMF) by EEMD method to obtain multichannel signals. Then, according to a cross correlation criterion, the corresponding IMF is selected as the input matrix of ICA. Finally, the compound faults can be separated effectively by executing ICA method, which makes the fault features more easily extracted and more clearly identified. Experimental results validate the effectiveness of the proposed method in compound fault separating, which works not only for the outer race defect, but also for the rollers defect and the unbalance fault of the experimental system.
77 citations
TL;DR: The use of RELICA is demonstrated on EEG data recorded from 14 subjects performing a working memory experiment and it is shown that many brain and ocular artifact ICs are correctly classified as "stable" (highly repeatable across decompositions of bootstrapped versions of the input data).
74 citations
TL;DR: The identification conditions for a general IVA formulation, which accounts for linear, nonlinear, and sample-to-sample dependencies, and performance bounds in terms of the Cramér-Rao lower bound are provided for demixing matrices and interference to source ratio.
Abstract: Recently, an extension of independent component analysis (ICA) from one to multiple datasets, termed independent vector analysis (IVA), has been a subject of significant research interest IVA has also been shown to be a generalization of Hotelling's canonical correlation analysis In this paper, we provide the identification conditions for a general IVA formulation, which accounts for linear, nonlinear, and sample-to-sample dependencies The identification conditions are a generalization of previous results for ICA and for IVA when samples are independently and identically distributed Furthermore, a principal aim of IVA is identification of dependent sources between datasets Thus, we provide additional conditions for when the arbitrary ordering of the estimated sources can be common across datasets Performance bounds in terms of the Cramer-Rao lower bound are also provided for demixing matrices and interference to source ratio The performance of two IVA algorithms are compared to the theoretical bounds
Journal Article•
TL;DR: An extension of slow feature analysis is presented and test as a novel approach to nonlinear blind source separation that relies on temporal correlations and iteratively reconstructs a set of statistically independent sources from arbitrary nonlinear instantaneous mixtures.
Abstract: We present and test an extension of slow feature analysis as a novel approach to nonlinear blind source separation. The algorithm relies on temporal correlations and iteratively reconstructs a set of statistically independent sources from arbitrary nonlinear instantaneous mixtures. Simulations show that it is able to invert a complicated nonlinear mixture of two audio signals with a high reliability. The algorithm is based on a mathematical analysis of slow feature analysis for the case of input data that are generated from statistically independent sources.
TL;DR: In this paper, a robust feature extraction scheme for the rolling element bearing (REB) fault diagnosis is proposed by combining the envelope extraction and the independent component analysis (ICA), which is not only utilized to obtain the impulsive component corresponding to the faults from the REB, but also to reduce the dimension of vibration sources included in the sensor-picked signals.
TL;DR: A new concept of multiple MPs (MMPs) is proposed, aimed at synthesizing the spectral-spatial information extracted from the multicomponent base images, and further enhancing the classification accuracy of MPs.
Abstract: Morphological profiles (MPs) are a useful tool for remotely sensed image classification. These profiles are constructed on a base image that can be a single band of a multicomponent remote sensing image. Principal component analysis (PCA) has been used to provide other base images to construct MPs in high-dimensional remote sensing scenes such as hyperspectral images [e.g., by deriving the first principal components (PCs) and building the MPs on the first few components]. In this paper, we discuss several strategies for producing the base images for MPs, and further categorize the considered methods into four classes: 1) linear, 2) nonlinear, 3) manifold learning-based, and 4) multilinear transformation-based. It is found that the multilinear PCA (MPCA) is a powerful approach for base image extraction. That is because it is a tensor-based feature representation approach, which is able to simultaneously exploit the spectral–spatial correlation between neighboring pixels. We also show that independent component analysis (ICA) is more effective for constructing base images than PCA. Another important contribution of this paper is a new concept of multiple MPs (MMPs), aimed at synthesizing the spectral–spatial information extracted from the multicomponent base images, and further enhancing the classification accuracy of MPs. Moreover, we propose two different strategies to interpret the newly proposed MMPs by considering their hyperdimensional feature space: 1) decision fusion and 2) sparse classifier based on multinomial logistic regression (MLR). Experiments conducted on three well-known hyperspectral datasets are used to quantitatively assess the accuracy of different algorithms.
01 Jan 2014
TL;DR: The results reveal that the proposed ICi-ensemble method outperformed the previous method using a single ICi with ~ 7% (91.6% versus 84.3%) in the cognitive state classification and favors the application of BCI in natural environments.
Abstract: Recently, successful applications of independent component analysis (ICA) to electroencephalographic (EEG) signals have yielded tremendous insights into brain processes that underlie human cognition Many studies have further established the feasibility of using independent processes to elucidate human cognitive states However, various technical problems arise in the building of an online brain-computer interface (BCI) These include the lack of an automatic procedure for selecting independent components of interest (ICi) and the potential risk of not obtaining a desired ICi Therefore, this study proposes an ICi-ensemble method that uses multiple classifiers with ICA processing to improve upon existing algorithms The mechanisms that are used in this ensemble system include: 1) automatic ICi selection; 2) extraction of features of the resultant ICi; 3) the construction of parallel pipelines for effectively training multiple classifiers; and a 4) simple process that combines the multiple decisions The proposed ICi-ensemble is demonstrated in a typical BCI application, which is the monitoring of participants' cognitive states in a realistic sustained-attention driving task The results reveal that the proposed ICi-ensemble outperformed the previous method using a single ICi with ~ 7% (916% versus 843%) in the cognitive state classification Additionally, the proposed ICi-ensemble method that characterizes the EEG dynamics of multiple brain areas favors the application of BCI in natural environments
TL;DR: Experimental results of simulated and measured data have demonstrated the effectiveness of the proposed method for NBI mitigation using the independent subspace analysis.
Abstract: The mitigation of narrow-band interference (NBI) is an appealing topic in the synthetic aperture radar (SAR) community. It is an underdetermined single-channel separation problem. This paper proposes a method for NBI mitigation using the independent subspace analysis. First, each single pulse is transformed onto a manifold time-frequency distribution by the short-time Fourier transform (STFT). Then, the singular value analysis is carried out to extract the prominent features corresponding to the NBIs. Next, independent component analysis is employed to obtain statistically independent basis components. Furthermore, the independent subspaces corresponding to NBI are reconstructed and subtracted from the raw signal space. The signal with NBI mitigated is resynthesized by inverse STFT. Finally, after processing all the pulses, a well-focused SAR imagery is obtained by a conventional imaging algorithm. Experimental results of simulated and measured data have demonstrated the effectiveness of the proposed method.
TL;DR: ICA was used as a blind source separation method on a Raman image of a pharmaceutical tablet and it was shown that under-decomposition of the matrix could lead to better signal quality (compared to the pure spectra) but in that case the contributions due to minor components or effects were not correctly identified and extracted.
TL;DR: In this article, various independent component functionals based on the fourth moments are discussed in detail, starting with the corresponding optimization problems, deriving the estimating equations and estimation algorithms, and finding asymptotic statistical properties of the estimates.
Abstract: In independent component analysis it is assumed that the components of the observed random vector are linear combinations of latent independent random variables, and the aim is then to find an estimate for a transformation matrix back to these independent components. In the engineering literature, there are several traditional estimation procedures based on the use of fourth moments, such as FOBI (fourth order blind identification), JADE (joint approximate diagonalization of eigenmatrices), and FastICA, but the statistical properties of these estimates are not well known. In this paper various independent component functionals based on the fourth moments are discussed in detail, starting with the corresponding optimization problems, deriving the estimating equations and estimation algorithms, and finding asymptotic statistical properties of the estimates. Comparisons of the asymptotic variances of the estimates in wide independent component models show that in most cases JADE and the symmetric version of FastICA perform better than their competitors.
TL;DR: Wang et al. as discussed by the authors proposed a fault detection method for non-Gaussian process based on robust independent component analysis (RobustICA), which can effectively reduce the effects of outliers.
TL;DR: The results obtained demonstrate that the proposed NoisyICAn-based monitoring method outperforms the conventional noise-free ICA- based monitoring methods as well as the benchmark monitoring methods based on the existing noisy ICA schemes adopted from blind source separation, in terms of the fault detection time and local fault detection rate.
TL;DR: The purpose of this paper is to describe the use of source separation for sEMG using ICA, and to demonstrate the actual use in practical s EMG experiments, when the number of recording channels for electrical muscle activities is varied.
Abstract: Surface electromyogram sEMG is a technique in which electrodes are placed on the skin overlying a muscle to detect the electrical activity. Multiple electrical sensors are essential for extracting intrinsic physiological and contextual information from the corresponding sEMG signals. The reason, why more than just one sEMG signal capture has to be used, is as follows: Due to signal propagation inside the human body in terms of an electrical conductor, there cannot be a one-to-one mapping of activities between muscle fibre groups and corresponding sEMG sensing electrodes. Each of such electrodes rather records a composition of many, and widely activity-independent signals, and such kind of raw signal capture cannot be efficiently used for pattern matching due to its linear dependency. On the other hand, Independent component analysis ICA provides the perfect answer of separating skin surface recordings into a set of independent muscle actions. Hence, there is a need for a method that indicates the quality of the sensor placements in sEMG. The purpose of this paper is to describe the use of source separation for sEMG using ICA. The actual use in practical sEMG experiments is demonstrated, when the number of recording channels for electrical muscle activities is varied.
TL;DR: This work proposes a simple but effective action recognition framework based on the recently proposed overcomplete ICA model that outperforms several state-of-the-art works on action recognition.
Book•
28 Jan 2014
TL;DR: This book offers a general overview of the basics of Blind Source Separation, important solutions and algorithms, and in-depth coverage of applications in image feature extraction, remote sensing image fusion, mixed-pixel decomposition of SAR images, image object recognition fMRI medical image processing, geochemical and geophysical data mining, mineral resources prediction and geoanomalies information recognition.
Abstract: A systematic exploration of both classic and contemporary algorithms in blind source separation with practical case studies The book presents an overview of Blind Source Separation, a relatively new signal processing method. Due to the multidisciplinary nature of the subject, the book has been written so as to appeal to an audience from very different backgrounds. Basic mathematical skills (e.g. on matrix algebra and foundations of probability theory) are essential in order to understand the algorithms, although the book is written in an introductory, accessible style.This book offers a general overview of the basics of Blind Source Separation, important solutions and algorithms, and in-depth coverage of applications in image feature extraction, remote sensing image fusion, mixed-pixel decomposition of SAR images, image object recognition fMRI medical image processing, geochemical and geophysical data mining, mineral resources prediction and geoanomalies information recognition. Firstly, the background and theory basics of blind source separation are introduced, which provides the foundation for the following work. Matrix operation, foundations of probability theory and information theory basics are included here. There follows the fundamental mathematical model and fairly new but relatively established blind source separation algorithms, such as Independent Component Analysis (ICA) and its improved algorithms (Fast ICA, Maximum Likelihood ICA, Overcomplete ICA, Kernel ICA, Flexible ICA, Non-negative ICA, Constrained ICA, Optimised ICA). The last part of the book considers the very recent algorithms in BSS e.g. Sparse Component Analysis (SCA) and Non-negative Matrix Factorization (NMF). Meanwhile, in-depth cases are presented for each algorithm in order to help the reader understand the algorithm and its application field.A systematic exploration of both classic and contemporary algorithms in blind source separation with practical case studiesPresents new improved algorithms aimed at different applications, such as image feature extraction, remote sensing image fusion, mixed-pixel decomposition of SAR images, image object recognition, and MRI medical image processingWith applications in geochemical and geophysical data mining, mineral resources prediction and geoanomalies information recognitionWritten by an expert team with accredited innovations in blind source separation and its applications in natural scienceAccompanying website includes a software system providing codes for most of the algorithms mentioned in the book, enhancing the learning experienceEssential reading for postgraduate students and researchers engaged in the area of signal processing, data mining, image processing and recognition, information, geosciences, life sciences.
TL;DR: A new ICA-based monitoring scheme that integrates an ensemble learning strategy with Bayesian inference is presented that can provide improved performance regardless of how the dominant ICs are determined.
TL;DR: Without requiring any artifact reference channel, the combination of Infomax and ADJUST improves the classification performance more than the other methods for both examined filtering cutoffs, i.e., 4 Hz and 25 Hz.
TL;DR: While epoch-based methods have been routinely applied the less often considered decomposition methods were clearly superior and therefore their use seems advisable and therefore the application of either SSS or tSSS is mandatory in Elekta systems.
TL;DR: In this paper, the authors proposed a source-space independent component analysis (ICA) for separation, tomography, and time-course reconstruction of EEG and MEG source signals, based on the application of singular value decomposition and ICA on the neuroelectrical signals from all brain voxels obtained post minimum-variance beamforming of sensor-space EEG or MEG.
TL;DR: This paper presents a thorough study on the performances of different independent component analysis algorithms for the extraction of class-discriminant information in remote sensing hyperspectral image classification, and considers the three implementations of ICA that are most widely used in signal processing, namely Infomax, FastICA, and JADE.
Abstract: This paper presents a thorough study on the performances of different independent component analysis (ICA) algorithms for the extraction of class-discriminant information in remote sensing hyperspectral image classification. The study considers the three implementations of ICA that are most widely used in signal processing, namely Infomax, FastICA, and JADE. The analysis aims to address a number of important issues regarding the use of ICA in the RS domain. Three scenarios are considered and the performances of the ICA algorithms are evaluated and compared against each other, in order to reach the final goal of identifying the most suitable approach to the analysis of hyperspectral images in supervised classification. Different feature extraction and selection techniques are used for dimensionality reduction with ICA and are then compared to the commonly used strategy, which is based on preprocessing data with principal components analysis (PCA) prior to classification. Experimental results obtained on three real hyperspectral data sets from each of the considered algorithms are presented and analyzed in terms of both classification accuracies and computational time.
TL;DR: ICA deconvolution considerably improves the application range of direct NMR spectroscopy for analysis of complex mixtures and SIMPLISMA and MILCA algorithms were found to be preferable for NMRSpectroscopic mixture analysis and showed similar performance.
Abstract: The major challenge facing NMR spectroscopic mixture analysis is the overlapping of signals and the arising impossibility to easily recover the structures for identification of the individual components and to integrate separated signals for quantification. In this paper, various independent component analysis (ICA) algorithms [mutual information least dependent component analysis (MILCA); stochastic non-negative ICA (SNICA); joint approximate diagonalization of eigenmatrices (JADE); and robust, accurate, direct ICA algorithm (RADICAL)] as well as deconvolution methods [simple-to-use-interactive self-modeling mixture analysis (SIMPLISMA) and multivariate curve resolution-alternating least squares (MCR-ALS)] are applied for simultaneous (1)H NMR spectroscopic determination of organic substances in complex mixtures. Among others, we studied constituents of the following matrices: honey, soft drinks, and liquids used in electronic cigarettes. Good quality spectral resolution of up to eight-component mixtures was achieved (correlation coefficients between resolved and experimental spectra were not less than 0.90). In general, the relative errors in the recovered concentrations were below 12%. SIMPLISMA and MILCA algorithms were found to be preferable for NMR spectra deconvolution and showed similar performance. The proposed method was used for analysis of authentic samples. The resolved ICA concentrations match well with the results of reference gas chromatography-mass spectrometry as well as the MCR-ALS algorithm used for comparison. ICA deconvolution considerably improves the application range of direct NMR spectroscopy for analysis of complex mixtures.
04 May 2014
TL;DR: Both simulated and real fMRI results show that this method selects more useful ICA runs than those selected by the widely used ICASSO software and that it is a more objective and better motivated approach to evaluate results and hence a promising tool for ICA analysis of fMRI data.
Abstract: Independent component analysis (ICA) has proven quite useful for the analysis of functional magnetic resonance imaging (fMRI) data. However, stability of ICA decompositions is an issue in ICA of fMRI analysis primarily due to the noisy nature of fMRI data and the iterative nature of algorithms. In this work, we present an approach that utilizes an objective criterion and that is particularly suitable for image analysis to select the best of multiple ICA runs to use for further analysis and inference. In addition, a growing number of studies are focusing on the decomposition of single subject data and/or using high ICA model order, which both require an effective way to align components obtained from different ICA runs. In this paper, while presenting a method that provides superior performance in selecting the best run and interpreting the statistical reliability of ICA estimates, we also address the component sorting issue. Both simulated and real fMRI results show that our method selects more useful ICA runs than those selected by the widely used ICASSO software and that it is a more objective and better motivated approach to evaluate results and hence a promising tool for ICA analysis of fMRI data.