To see the other types of publications on this topic, follow the link: Independent component analysis.
Dissertations / Theses on the topic 'Independent component analysis'
Create a spot-on reference in APA, MLA, Chicago, Harvard, and other styles
Consult the top 50 dissertations / theses for your research on the topic 'Independent component analysis.'
Next to every source in the list of references, there is an 'Add to bibliography' button. Press on it, and we will generate automatically the bibliographic reference to the chosen work in the citation style you need: APA, MLA, Harvard, Chicago, Vancouver, etc.
You can also download the full text of the academic publication as pdf and read online its abstract whenever available in the metadata.
Browse dissertations / theses on a wide variety of disciplines and organise your bibliography correctly.
1
Gao, Pei. "Nonlinear independent component analysis." Thesis, University of Newcastle Upon Tyne, 2007. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.437979.
Full text
2
Harmeling, Stefan. "Independent component analysis and beyond." Phd thesis, [S.l. : s.n.], 2004. http://deposit.ddb.de/cgi-bin/dokserv?idn=973631805.
Full text
3
Blaschke, Tobias. "Independent component analysis and slow feature analysis." Doctoral thesis, Humboldt-Universität zu Berlin, Mathematisch-Naturwissenschaftliche Fakultät I, 2005. http://dx.doi.org/10.18452/15270.
Full textAbstract:
Der Fokus dieser Dissertation liegt auf den Verbindungen zwischen ICA (Independent Component Analysis - Unabhängige Komponenten Analyse) und SFA (Slow Feature Analysis - Langsame Eigenschaften Analyse). Um einen Vergleich zwischen beiden Methoden zu ermöglichen wird CuBICA2, ein ICA Algorithmus basierend nur auf Statistik zweiter Ordnung, d.h. Kreuzkorrelationen, vorgestellt. Dieses Verfahren minimiert zeitverzögerte Korrelationen zwischen Signalkomponenten, um die statistische Abhängigkeit zwischen denselben zu reduzieren. Zusätzlich wird eine alternative SFA-Formulierung vorgestellt, die mit CuBICA2 verglichen werden kann. Im Falle linearer Gemische sind beide Methoden äquivalent falls nur eine einzige Zeitverzögerung berücksichtigt wird. Dieser Vergleich kann allerdings nicht auf mehrere Zeitverzögerungen erweitert werden. Für ICA lässt sich zwar eine einfache Erweiterung herleiten, aber ein ähnliche SFA-Erweiterung kann nicht im originären SFA-Sinne (SFA extrahiert die am langsamsten variierenden Signalkomponenten aus einem gegebenen Eingangssignal) interpretiert werden. Allerdings kann eine im SFA-Sinne sinnvolle Erweiterung hergeleitet werden, welche die enge Verbindung zwischen der Langsamkeit eines Signales (SFA) und der zeitlichen Vorhersehbarkeit desselben verdeutlich. Im Weiteren wird CuBICA2 und SFA kombiniert. Das Resultat kann aus zwei Perspektiven interpretiert werden. Vom ICA-Standpunkt aus führt die Kombination von CuBICA2 und SFA zu einem Algorithmus, der das Problem der nichtlinearen blinden Signalquellentrennung löst. Vom SFA-Standpunkt aus ist die Kombination eine Erweiterung der standard SFA. Die standard SFA extrahiert langsam variierende Signalkomponenten die untereinander unkorreliert sind, dass heißt statistisch unabhängig bis zur zweiten Ordnung. Die Integration von ICA führt nun zu Signalkomponenten die mehr oder weniger statistisch unabhängig sind.<br>Within this thesis, we focus on the relation between independent component analysis (ICA) and slow feature analysis (SFA). To allow a comparison between both methods we introduce CuBICA2, an ICA algorithm based on second-order statistics only, i.e.\ cross-correlations. In contrast to algorithms based on higher-order statistics not only instantaneous cross-correlations but also time-delayed cross correlations are considered for minimization. CuBICA2 requires signal components with auto-correlation like in SFA, and has the ability to separate source signal components that have a Gaussian distribution. Furthermore, we derive an alternative formulation of the SFA objective function and compare it with that of CuBICA2. In the case of a linear mixture the two methods are equivalent if a single time delay is taken into account. The comparison can not be extended to the case of several time delays. For ICA a straightforward extension can be derived, but a similar extension to SFA yields an objective function that can not be interpreted in the sense of SFA. However, a useful extension in the sense of SFA to more than one time delay can be derived. This extended SFA reveals the close connection between the slowness objective of SFA and temporal predictability. Furthermore, we combine CuBICA2 and SFA. The result can be interpreted from two perspectives. From the ICA point of view the combination leads to an algorithm that solves the nonlinear blind source separation problem. From the SFA point of view the combination of ICA and SFA is an extension to SFA in terms of statistical independence. Standard SFA extracts slowly varying signal components that are uncorrelated meaning they are statistically independent up to second-order. The integration of ICA leads to signal components that are more or less statistically independent.
4
Brock, James L. "Acoustic classification using independent component analysis /." Link to online version, 2006. https://ritdml.rit.edu/dspace/handle/1850/2067.
Full text
5
Papathanassiou, Christos. "Independent component analysis of magnetoencephalographic signals." Thesis, University of Surrey, 2003. http://epubs.surrey.ac.uk/771941/.
Full textAbstract:
Magnetoencephalography (MEG) is a non-invasive brain imaging technique which allows instant tracking of changes in brain activity. However, it is affected by strong artefact signals generated by the heart or the eye blinking. The blind source separation problem is typically encountered in MEG studies when a set of unknown signals, originating from different sources inside or outside the brain, is mixed with an also unknown mixing matrix during their recording. Independent component analysis (ICA) is a recently developed technique which aims to estimate the original sources given only the observed mixtures. ICA can decompose the observed data into the original biological sources. However, ICA suffers from a major intrinsic ambiguity. In particular, it cannot determine the order of extraction of the source signals. Thus, if there are numerous source signals hidden in lengthy MEG recordings, the extraction of the biological signal of interest can be an extremely prolonged procedure. In this thesis, a modification of the ordinary ICA is introduced in order to cope with this ambiguity. In case there is prior knowledge concerning one of the original signals, this information is exploited by adding a penalty/constraint term to the standard ICA quality function in order to favour the extraction of that particular signal. Our approach requires no reference signal, but the knowledge of some statistical property of one of the original sources, namely its autocorrelation function. Our algorithm is validated with simulated data for which the mixing matrix is known, and is also applied to real MEG data to remove artefact signals. Finally, it is demonstrated how ICA can simplify the ill-posed problem of localising the sources/dipoles in the cortex (inverse problem). The advantage of ICA lies in using nonaveraged trials. In addition, there is no need to know in advance the number of dipoles.
6
Miskin, James William. "Ensemble learning for independent component analysis." Thesis, University of Cambridge, 2001. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.621116.
Full text
7
Garvey, Jennie Hill. "Independent component analysis by entropy maximization (infomax)." Thesis, Monterey, Calif. : Naval Postgraduate School, 2007. http://bosun.nps.edu/uhtbin/hyperion-image.exe/07Jun%5FGarvey.pdf.
Full textAbstract:
Thesis (M.S. in Electrical Engineering)--Naval Postgraduate School, June 2007.<br>Thesis Advisor(s): Frank E. Kragh. "June 2007." Includes bibliographical references (p. 103). Also available in print.
8
Mitianoudis, Nikolaos. "Audio source separation using independent component analysis." Thesis, Queen Mary, University of London, 2004. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.406171.
Full text
9
Choudrey, Rizwan A. "Variational methods for Bayesian independent component analysis." Thesis, University of Oxford, 2002. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.275566.
Full text
10
Kalkan, Olcay Altınkaya Mustafa Aziz. "Independent component analysis applications in CDMA systems/." [s.l.]: [s.n.], 2004. http://library.iyte.edu.tr/tezler/master/elektronikvehaberlesme/T000473.rar.
Full text
11
Björling, Robin. "Denoising of Infrared Images Using Independent Component Analysis." Thesis, Linköping University, Department of Electrical Engineering, 2005. http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-4954.
Full textAbstract:
<p>Denna uppsats syftar till att undersöka användbarheten av metoden Independent Component Analysis (ICA) för brusreducering av bilder tagna av infraröda kameror. Speciellt fokus ligger på att reducera additivt brus. Bruset delas upp i två delar, det Gaussiska bruset samt det sensorspecifika mönsterbruset. För att reducera det Gaussiska bruset används en populär metod kallad sparse code shrinkage som bygger på ICA. En ny metod, även den byggandes på ICA, utvecklas för att reducera mönsterbrus. För varje sensor utförs, i den nya metoden, en analys av bilddata för att manuellt identifiera typiska mönsterbruskomponenter. Dessa komponenter används därefter för att reducera mönsterbruset i bilder tagna av den aktuella sensorn. Det visas att metoderna ger goda resultat på infraröda bilder. Algoritmerna testas både på syntetiska såväl som på verkliga bilder och resultat presenteras och jämförs med andra algoritmer.</p><br><p>The purpose of this thesis is to evaluate the applicability of the method Independent Component Analysis (ICA) for noise reduction of infrared images. The focus lies on reducing the additive uncorrelated noise and the sensor specific additive Fixed Pattern Noise (FPN). The well known method sparse code shrinkage, in combination with ICA, is applied to reduce the uncorrelated noise degrading infrared images. The result is compared to an adaptive Wiener filter. A novel method, also based on ICA, for reducing FPN is developed. An independent component analysis is made on images from an infrared sensor and typical fixed pattern noise components are manually identified. The identified components are used to fast and effectively reduce the FPN in images taken by the specific sensor. It is shown that both the FPN reduction algorithm and the sparse code shrinkage method work well for infrared images. The algorithms are tested on synthetic as well as on real images and the performance is measured.</p>
12
Sahambi, Harkirat S. "Appearance based object recognition using independent component analysis." Thesis, National Library of Canada = Bibliothèque nationale du Canada, 2000. http://www.collectionscanada.ca/obj/s4/f2/dsk1/tape4/PQDD_0017/MQ54320.pdf.
Full text
13
Wu, Hao-cun. "Independent component analysis and its applications in finance." Click to view the E-thesis via HKUTO, 2007. http://sunzi.lib.hku.hk/HKUTO/record/B39559099.
Full text
14
Beckmann, Christian F. "Independent component analysis for functional magnetic resonance imaging." Thesis, University of Oxford, 2004. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.404108.
Full text
15
吳浩存 and Hao-cun Wu. "Independent component analysis and its applications in finance." Thesis, The University of Hong Kong (Pokfulam, Hong Kong), 2007. http://hub.hku.hk/bib/B39559099.
Full text
16
Fiore, Ugo. "Improving Network Anomaly Detection with Independent Component Analysis." Doctoral thesis, Universita degli studi di Salerno, 2015. http://hdl.handle.net/10556/1978.
Full textAbstract:
2013 - 2014<br>Complexity, sophistication, and rate of growth of modern networks, coupled with the depth, continuity, and pervasiveness of their role in our everyday lives, stress the importance of identifying potential misuse or threats that could undermine regular operation.
To ensure an adequate and prompt reaction, anomalies in network traffic should be detected, classified, and identified as quickly and correctly as possible.
Several approaches focus on inspecting the content of packets traveling through the network, while other techniques aim at detecting suspicious activity by measuring the network state and comparing it with an expected baseline.
Formalizing a model for normal behavior requires the collection and analysis of traffic, in order to isolate a set of features capable of describing traffic completely and in a compact way.
The main focus of this dissertation is the quest for good representations for network traffic, representation that are abstract and can capture and describe much of the intricate structure of observed data in a simple manner.
In this way, some of the hidden factors and variables governing the traffic data generation process can be unveiled and disentangled and anomalous events can be spotted more reliably.
We adopted several methods to achieve such simpler representations, including Independent Component Analysis and deep learning architectures.
Machine learning techniques have been used for verifying the improvement in classification effectiveness that can be achieved with the proposed representations. [edited by Author]<br>XIII n.s.
17
Robila, Stefan Alexandru. "Independent component analysis based feature extraction for hyperspectral images." Related electronic resource: Current Research at SU : database of SU dissertations, recent titles available full text, 2002. http://wwwlib.umi.com/cr/syr/main.
Full text
18
Baylor, Martha-Elizabeth. "Analog optoelectronic independent component analysis for radio frequency signals." Connect to online resource, 2007. http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqdiss&rft_dat=xri:pqdiss:3288865.
Full text
19
E, Okwelume Gozie, and Ezeude Anayo Kingsley. "BLIND SOURCE SEPARATION USING FREQUENCY DOMAIN INDEPENDENT COMPONENT ANALYSIS." Thesis, Blekinge Tekniska Högskola, Avdelningen för signalbehandling, 2007. http://urn.kb.se/resolve?urn=urn:nbn:se:bth-1312.
Full textAbstract:
Our thesis work focuses on Frequency-domain Blind Source Separation (BSS) in which the received mixed signals are converted into the frequency domain and Independent Component Analysis (ICA) is applied to instantaneous mixtures at each frequency bin. Computational complexity is also reduced by using this method. We also investigate the famous problem associated with Frequency-Domain Blind Source Separation using ICA referred to as the Permutation and Scaling ambiguities, using methods proposed by some researchers. This is our main target in this project; to solve the permutation and scaling ambiguities in real time applications<br>Gozie: modebelu2001@yahoo.com Anayo: ezeudea@yahoo.com
20
Herrmann, Frank. "Independent component analysis with applications to blind source separation." Thesis, University of Liverpool, 2000. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.399147.
Full text
21
Rezaee, Sayed Majid. "A study on independent component analysis over galois fields." reponame:Repositório Institucional da UnB, 2015. http://dx.doi.org/10.26512/2015.12.D.20421.
Full textAbstract:
Dissertação (mestrado)—Universidade de Brasília, Faculdade de Tecnologia, Departamento de Engenharia Elétrica, 2015.<br>Submitted by Fernanda Percia França (fernandafranca@bce.unb.br) on 2016-02-22T20:16:00Z
No. of bitstreams: 1
2015_SayedMajidRezaee.pdf: 1299294 bytes, checksum: 5ae8992f8de2423bc23cf06bdeaeeb09 (MD5)<br>Approved for entry into archive by Marília Freitas(marilia@bce.unb.br) on 2016-05-26T16:32:15Z (GMT) No. of bitstreams: 1
2015_SayedMajidRezaee.pdf: 1299294 bytes, checksum: 5ae8992f8de2423bc23cf06bdeaeeb09 (MD5)<br>Made available in DSpace on 2016-05-26T16:32:15Z (GMT). No. of bitstreams: 1
2015_SayedMajidRezaee.pdf: 1299294 bytes, checksum: 5ae8992f8de2423bc23cf06bdeaeeb09 (MD5)<br>Nas últimas décadas, o problema de separação cega de fontes (BSS, do inglês Blind Source Separation) – que trata de estimar um conjunto desconhecido de fontes de sinais a partir de versões misturadas destes – tornou-se relevante em vários campos da engenharia, incluindo o processamento matricial, comunicações sem fio, processamento de sinais médicos, processamento de voz e engenharia biomédica. A fim de resolver o problema de BSS no contexto de modelos lineares, considerando-se várias técnicas possíveis, a Análise de Componentes Independentes (ICA, do inglês Independent Component Analysis) – que utiliza a independência estatística das fontes como uma premissa – demonstrou ser uma das mais importantes estratégias de solução. Além disso, embora o modelo de BSS/ICA para sinais reais ou complexos esteja bem estabelecido, a recente perspectiva de uma formulação do problema com sinais e modelos definidos em corpos de Galois oferece várias possibilidades de análise e contribuições. Esta dissertação de mestrado realiza um estudo da Análise de Componentes Independentes em corpos de Galois, considerando os conceitos teóricos e abordagens para o problema, assim como dos algoritmos estado-da-arte até agora propostos, em termos de suas capacidades de separação e custo computacional. Especificamente, as técnicas dos algoritmos AMERICA e MEXICO são estudadas juntamente com o algoritmo cobICA. Como as simulações experimentais indicam, devido à sua complexidade computacional menor e uma qualidade de desempenho satisfatório, o algoritmo cobICA apresenta-se como uma solução de compromisso entre os algoritmos AMERICA e MEXICO para executar BSS/ICA em corpos de Galois.<br>Over the past decades, the Blind Source Separation (BSS) problem – which deals with estimating an unknown set of source signals from their measured mixtures –has become prevalent in several engineering fields, including array processing, wireless communications, medical signal processing, speech processing and biomedical engineering. In order to solve the BSS problem in the context of linear models, considering several possible techniques, Independent Component Analysis (ICA) – which uses statistical independence of the source signals as a premise – has been shown to be one of the most important approaches. Furthermore, although the BSS/ICA framework for real- or complex-valued signals is firmly established, the recent perspective of a BSS/ICA formulation where the signals and models are defined over Galois fields gives several possibilities of analyzes and contributions. This Master’s thesis performs a study on Independent Component Analysis over Galois fields, considering the theoretical concepts and aspects of the problem and the investigation, in terms of capability and efficiency, of the state-of-the-art algorithms so far introduced. In this context, AMERICA and MEXICO techniques are studied, along with cobICA algorithm – a bioinspired framework based on cob-aiNet[C] immune-inspired algorithm –, mainly focusing on comparing the quality of separation and on discussing the computational burden of each technique. As the experimental simulations indicate, due to its lower computational complexity and a satisfactory performance quality, cobICA takes place as a compromise solution between AMERICA and MEXICO algorithms, to perform BSS/ICA over Galois fields.
22
Shawli, Alaa. "Scoring the SF-36 health survey in scleroderma using independent component analysis and principle component analysis." Thesis, McGill University, 2011. http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=97180.
Full textAbstract:
The short form SF-36 survey is a widely used survey of patient health related quality of life. It yields eight subscale scores of functional health and well-being that are summarized by two physical and mental component summary scores. However, recent studies have reported inconsistent results between the eight subscales and the two component summary measures when the scores are from a sick population. They claim that this problem is due to the method used to compute the SF-36 component summary scores, which is based on principal component analysis with orthogonal rotation.In this thesis, we explore various methods in order to identify a method that is more accurate in obtaining the SF-36 physical and mental component component summary scores (PCS and MCS), with a focus on diseased patient subpopulations. We first explore traditional data analysis methods such as principal component analysis (PCA) and factor analysis using maximum likelihoodestimation and apply orthogonal and oblique rotations with both methods to data from the Canadian Scleroderma Research Group registry. We compare these common approaches to a recently developed data analysis method from signal processing and neural network research, independent component analysis (ICA). We found that oblique rotation is the only method that reduces the meanmental component scores to best match the mental subscale scores. In order to try to better elucidate the differences between the orthogonal and oblique rotation, we studied the performance of PCA with the two approaches for recovering the true physical and mental component summary scores in a simulated diseased population where we knew the truth. We explored the methods in situations where the true scores were independent and when they were also correlated. We found that ICA and PCA with orthogonal rotation performed very similarly when the data were generated to be independent, but differently (with ICA performing worse) when the data were generated to be correlated. PCA with oblique rotation tended to perform worse than both methods when the data were independent, but better when the data were correlated. We also discuss the connection between ICA and PCA with orthogonal rotation, which lends strength to the use of the varimax rotation for the SF-36.Finally, we applied ICA to the scleroderma data and found relatively low correlation between ICA and unrotated PCA in estimating the PCS and MCS scores and very high correlation between ICA and PCA with varimax rotation. PCA with oblique rotation also had a relatively high correlation with ICA. Hence, we concluded that ICA could be seen as a compromise solution between the two methods.<br>La version abrégée du questionnaire SF-36 est largement utilisée pour valider la qualité de vie reliée à la santé. Ce questionnaire fournit huit scores s'attardant à la capacité fonctionnelle et au bien-être, lesquels sont regroupés en cotes sommaires attribuées aux composantes physiques et mentales. Cependant, des études récentes ont rapporté des résultats contradictoires entre les huit sous-échelles et les deux cotes sommaires lorsque les scores sont obtenus auprès de sujets malades. Cette discordance serait due à la méthode utilisée pour calculer les cotes sommaires du SF-36 qui est fondée sur l'analyse en composantes principales avec rotation orthogonale.Dans cette thèse, nous explorons diverses méthodes dans le but d'identifier une méthode plus précise pour calculer les cotes sommaires du SF-36 attribuées aux composantes physiques et mentales (CCP et CCM), en mettant l'accent sur des sous-populations de sujets malades. Nous évaluerons d'abord des méthodes traditionnelles d'analyse de données, telles que l'analyse en composantes principales (ACP) et l'analyse factorielle, en utilisant l'étude de l'estimation du maximum de vraisemblance et en appliquant les rotations orthogonale et oblique aux deux méthodes sur les données du registre du Groupe de recherche canadien sur la sclérodermie. Nous comparons ces approches courantes à une méthode d'analyse de données développée récemment à partir de travaux de recherche sur le réseau neuronal et le traitement du signal, l'analyse en composantes indépendantes (ACI).Nous avons découvert que la rotation oblique est la seule méthode qui réduit les cotes attribuées aux composantes mentales moyennes afin de mieux les corréler aux scores de la sous-échelle des symptômes mentaux. Dans le but de mieux comprendre les différences entre la rotation orthogonale et la rotation oblique, nous avons étudié le rendement de l'ACP avec deux approches pour déterminer les véritables cotes sommaires attribuées aux composantes physiques et mentales dans une population simulée de sujets malades pour laquelle les données étaient connues. Nous avons exploré les méthodes dans des situations où les scores véritables étaient indépendants et lorsqu'ils étaient corrélés. Nous avons conclu que le rendement de l'ACI et de l'ACP associées à la rotation orthogonale était très similaire lorsque les données étaient indépendantes, mais que le rendement différait lorsque les données étaient corrélées (ACI étant moins performante). L'ACP associée à la rotation oblique a tendance à être moins performante que les deux méthodes lorsque les données étaient indépendantes, mais elle est plus performante lorsque les données étaient corrélées. Nous discutons également du lien entre l'ACI et l'ACP avec la rotation orthogonale, ce qui appuie l'emploi de la rotation varimax dans le questionnaire SF 36.Enfin, nous avons appliqué l'ACI aux données sur la sclérodermie et nous avons mis en évidence une corrélation relativement faible entre l'ACI et l'ACP sans rotation dans l'estimation des scores CCP et CCM, et une corrélation très élevée entre l'ACI et l'ACP avec rotation varimax. L'ACP avec rotation oblique présentait également une corrélation relativement élevée avec l'ACI. Par conséquent, nous en avons conclu que l'ACI pourrait servir de solution de compromis entre ces deux méthodes.
23
Sohr, Mandy. "Analysis of functional magnetic resonance imaging time series by independent component analysis." [S.l.] : [s.n.], 2007. http://deposit.ddb.de/cgi-bin/dokserv?idn=985601965.
Full text
24
Afsari, Bijan. "Gradient flow based matrix joint diagonalization for independent component analysis." College Park, Md. : University of Maryland, 2004. http://hdl.handle.net/1903/1352.
Full textAbstract:
Thesis (M.S.) -- University of Maryland, College Park, 2004.<br>Thesis research directed by: Electrical Engineering. Title from t.p. of PDF. Includes bibliographical references. Published by UMI Dissertation Services, Ann Arbor, Mich. Also available in paper.
25
Elsabrouty, Maha. "Riemannian geometry based blind signal separation using independent component analysis." Thesis, University of Ottawa (Canada), 2006. http://hdl.handle.net/10393/29291.
Full textAbstract:
Blind Source Separation is one of the newest and most active research areas in adaptive filtering. It represents the solution for many real situations in the audio, speech processing and telecommunication fields. The word "blind" reflects the fact that neither the source nor the mixing channel is known. This is, clearly, a more difficult situation compared to conventional adaptive filtering problems. Algorithms developed for blind separation reflect this difficulty. They possess a higher degree of sophistication compared with algorithms in other adaptive filtering approaches. The cost functions employed for blind separation are mostly based, implicitly or explicitly, on higher order statistics, while some of the adaptive algorithms developed for blind signal separation witnessed the introduction of the powerful principle of differential geometry to modify the LMS-based algorithms to what is known as the natural-gradient update.
The aim of this work is to generalize differential geometry algorithms for different cost functions of blind signal separation and provide new faster converging RLS-based and Newton-based algorithms using the natural gradient update. The thesis starts by providing a thorough review of the existing solutions developed in the literature for blind separation. This is followed by a study of the mathematical basis of Riemannian geometry and providing an engineering insight into the intrinsic geometry of curved spaces and its relation to optimization in adaptive filtering. Several new update algorithms are then developed throughout the thesis. They are structured to perform at faster convergence rates even in difficult mixing situations. These algorithms provide significantly improved performance in comparison with existing algorithms and are suitable for on-line applications.
26
Sarfraz, M. "Role of independent component analysis in intelligent ECG signal processing." Thesis, University of Salford, 2014. http://usir.salford.ac.uk/33200/.
Full textAbstract:
The Electrocardiogram (ECG) reflects the activities and the attributes of the human heart and reveals very important hidden information in its structure. The information is extracted by means of ECG signal analysis to gain insights that are very crucial in explaining and identifying various pathological conditions. The feature extraction process can be accomplished directly by an expert through, visual inspection of ECGs printed on paper or displayed on a screen. However, the complexity and the time taken for the ECG signals to be visually inspected and manually analysed means that it‟s a very tedious task thus yielding limited descriptions. In addition, a manual ECG analysis is always prone to errors: human oversights. Moreover ECG signal processing has become a prevalent and effective tool for research and clinical practices. A typical computer based ECG analysis system includes a signal preprocessing, beats detection and feature extraction stages, followed by classification. Automatic identification of arrhythmias from the ECG is one important biomedical application of pattern recognition. This thesis focuses on ECG signal processing using Independent Component Analysis (ICA), which has received increasing attention as a signal conditioning and feature extraction technique for biomedical application. Long term ECG monitoring is often required to reliably identify the arrhythmia. Motion induced artefacts are particularly common in ambulatory and Holter recordings, which are difficult to remove with conventional filters due to their similarity to the shape of ectopic xiii beats. Feature selection has always been an important step towards more accurate, reliable and speedy pattern recognition. Better feature spaces are also sought after in ECG pattern recognition applications. Two new algorithms are proposed, developed and validated in this thesis, one for removing non-trivial noises in ECGs using the ICA and the other deploys the ICA extracted features to improve recognition of arrhythmias. Firstly, independent component analysis has been studied and found effective in this PhD project to separate out motion induced artefacts in ECGs, the independent component corresponding to noise is then removed from the ECG according to kurtosis and correlation measurement. The second algorithm has been developed for ECG feature extraction, in which the independent component analysis has been used to obtain a set of features, or basis functions of the ECG signals generated hypothetically by different parts of the heart during the normal and arrhythmic cardiac cycle. ECGs are then classified based on the basis functions along with other time domain features. The selection of the appropriate feature set for classifier has been found important for better performance and quicker response. Artificial neural networks based pattern recognition engines are used to perform final classification to measure the performance of ICA extracted features and effectiveness of the ICA based artefacts reduction algorithm. The motion artefacts are effectively removed from the ECG signal which is shown by beat detection on noisy and cleaned ECG signals after ICA processing. Using the ICA extracted feature sets classification of ECG arrhythmia into eight classes with fewer independent components and very high classification accuracy is achieved.
27
Zakeri, Zohreh. "Optimised use of independent component analysis for EEG signal processing." Thesis, University of Birmingham, 2017. http://etheses.bham.ac.uk//id/eprint/7430/.
Full textAbstract:
Electroencephalography (EEG) is the prevalent technique for monitoring brain function. It employs a set of electrodes on the scalp to measure the electrical activity of the brain. EEG is mainly used by researchers to study the brain’s responses to a specific stimulus - the event-related potentials (ERPs). Different types of unwanted signals, which are known as artefacts, usually mix with the EEG at any point during the recording process. As the amplitudes of the EEG and ERPs are very small (in the order of microvolts), they can be buried in the artefacts which have very high amplitudes in the order of millivolts. Therefore, contamination of EEG activity by the artefacts can degrade the quality of the EEG recording and may cause error in EEG/ERP signal interpretation. Several EEG artefact removal methods already exist in the literature and these previous studies have concentrated on manual or automatic detection of either one or, of a few types of EEG artefacts. Among the proposed methods, Independent Component Analysis (ICA) based techniques are commonly applied to successfully detect the artefacts. Different types of ICA algorithms have been developed, which aim to estimate the individual sources of a linearly mixed signal. However, the estimation criterion differs across various ICA algorithms, which may deliver different results.
28
Wang, Suogang. "Enhancing brain-computer interfacing through advanced independent component analysis techniques." Thesis, University of Southampton, 2009. https://eprints.soton.ac.uk/65897/.
Full textAbstract:
A brain-computer interface (BCI) is a direct communication system between a brain and an external device in which messages or commands sent by an individual do not pass through the brain’s normal output pathways but is detected through brain signals. Some severe motor impairments, such as Amyothrophic Lateral Sclerosis, head trauma, spinal injuries and other diseases may cause the patients to lose their muscle control and become unable to communicate with the outside environment. Currently no effective cure or treatment has yet been found for these diseases. Therefore using a BCI system to rebuild the communication pathway becomes a possible alternative solution. Among different types of BCIs, an electroencephalogram (EEG) based BCI is becoming a popular system due to EEG’s fine temporal resolution, ease of use, portability and low set-up cost. However EEG’s susceptibility to noise is a major issue to develop a robust BCI. Signal processing techniques such as coherent averaging, filtering, FFT and AR modelling, etc. are used to reduce the noise and extract components of interest. However these methods process the data on the observed mixture domain which mixes components of interest and noise. Such a limitation means that extracted EEG signals possibly still contain the noise residue or coarsely that the removed noise also contains part of EEG signals embedded. Independent Component Analysis (ICA), a Blind Source Separation (BSS) technique, is able to extract relevant information within noisy signals and separate the fundamental sources into the independent components (ICs). The most common assumption of ICA method is that the source signals are unknown and statistically independent. Through this assumption, ICA is able to recover the source signals. Since the ICA concepts appeared in the fields of neural networks and signal processing in the 1980s, many ICA applications in telecommunications, biomedical data analysis, feature extraction, speech separation, time-series analysis and data mining have been reported in the literature. In this thesis several ICA techniques are proposed to optimize two major issues for BCI applications: reducing the recording time needed in order to speed up the signal processing and reducing the number of recording channels whilst improving the final classification performance or at least with it remaining the same as the current performance. These will make BCI a more practical prospect for everyday use. This thesis first defines BCI and the diverse BCI models based on different control patterns. After the general idea of ICA is introduced along with some modifications to ICA, several new ICA approaches are proposed. The practical work in this thesis starts with the preliminary analyses on the Southampton BCI pilot datasets starting with basic and then advanced signal processing techniques. The proposed ICA techniques are then presented using a multi-channel event related potential (ERP) based BCI. Next, the ICA algorithm is applied to a multi-channel spontaneous activity based BCI. The final ICA approach aims to examine the possibility of using ICA based on just one or a few channel recordings on an ERP based BCI. The novel ICA approaches for BCI systems presented in this thesis show that ICA is able to accurately and repeatedly extract the relevant information buried within noisy signals and the signal quality is enhanced so that even a simple classifier can achieve good classification accuracy. In the ERP based BCI application, after multichannel ICA the data just applied to eight averages/epochs can achieve 83.9% classification accuracy whilst the data by coherent averaging can reach only 32.3% accuracy. In the spontaneous activity based BCI, the use of the multi-channel ICA algorithm can effectively extract discriminatory information from two types of singletrial EEG data. The classification accuracy is improved by about 25%, on average, compared to the performance on the unpreprocessed data. The single channel ICA technique on the ERP based BCI produces much better results than results using the lowpass filter. Whereas the appropriate number of averages improves the signal to noise rate of P300 activities which helps to achieve a better classification. These advantages will lead to a reliable and practical BCI for use outside of the clinical laboratory.
29
Vigon, Laurence Celine. "Independent component analysis techniques and their performance evaluation for electroencephalography." Thesis, Sheffield Hallam University, 2002. http://shura.shu.ac.uk/20479/.
Full textAbstract:
The ongoing electrical activity of the brain is known as the electroencephalogram (EEG). Evoked potentials (EPs) are voltage deviations in the EEG elicited in association with stimuli. EPs provide clinical information by allowing an insight into neurological processes. The amplitude of EPs is typically several times less than the background EEG. The background EEG has the effect of obscuring the EPs and therefore appropriate signal processing is required for their recovery. The EEG waveforms recorded from electrodes placed on the scalp contains the ongoing background EEG, EPs from various brain sources as well as signal components with sources external to the brain. An example of externally generated signal which is picked up by the electrodes on the scalp is the electrooculogram (EOG). This signal is generated by the eyes when eye movements or blinks are performed. Saccade-related EEG waveforms were recorded from 7 normal subjects. A signal source separation technique, namely the independent component analysis (ICA) algorithm of Bell and Sejnowski (hereafter refereed to as BS_ICA), was employed to analyse the recorded waveforms. The effectiveness of the BS_ICA algorithm as well as that of the ICA algorithm of Cardoso, was investigated for removing ocular artefact (OA) from the EEG. It was quantitavely demonstrated that both ICA algorithms were more effective than the conventional correlation-based techniques for removing the OA from the EEG.A novel iterative synchronised averaging method for EPs was devised. The method optimally synchronised the waveforms from successive trials with respect to the event of interest prior to averaging and thus preserved the features of the signals components that were time-locked to the event. The recorded EEG waveforms were analysed using BS_ICA and saccade-related components (frontal and occipital pre-saccadic potentials, and the lambda wave) were extracted and their scalp topographies were obtained. This initial study highlighted some limitations of the conventional ICA approach of Bell and Sejnowski for analysing saccade-related EEG waveforms. Novel techniques were devised in order to improve the performance of the ICA algorithm of Bell and Sejnowski for extracting the lambda wave EP component. One approach involved designing a template-model that represented the temporal characteristics of a lambda wave. Its incorporation into the BS_ICA algorithm improved the signal source separation ability of the algorithm for extracting the lambda wave from the EEG waveforms. The second approach increased the effective length of the recorded EEG traces prior to their processing by the BS_ICA algorithm. This involved abutting EEG traces from an appropriate number of successive trials (a trial was a set of waveforms recorded from 64 electrode locations in a experiment involving a saccade performance). It was quantitatively demonstrated that the process of abutting EEG waveforms was a valuable pre-processing operation for the ICA algorithm of Bell and Sejnowski when extracting the lambda wave. A Fuzzy logic method was implemented to identify BS_ICA-extracted single-trial saccade-related lambda waves. The method provided an effective means to automate the identification of the lambda waves extracted by BS_ICA. The approach correctly identified the single-trial lambda waves with an Accuracy of 97.4%.
30
Kent, Michael. "The value of independent component analysis in identifying climate processes." Master's thesis, University of Cape Town, 2011. http://hdl.handle.net/11427/11457.
Full textAbstract:
We find that ICA could make a useful contribution through the identification of the land based seasonal cycle and the ocean based seasonal cycle. These qualities mean that ICA may further prove a useful tool in the problem of identifying components of climate change signals from ensembles of multiple climate models.
31
Johnson, Robert Spencer. "Incorporation of prior information into independent component analysis of FMRI." Thesis, University of Oxford, 2012. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.711637.
Full text
32
Wang, Fei. "Vertical beam emittance correction with independent component analysis measurement method." [Bloomington, Ind.] : Indiana University, 2008. http://gateway.proquest.com/openurl?url_ver=Z39.88-2004&rft_val_fmt=info:ofi/fmt:kev:mtx:dissertation&res_dat=xri:pqdiss&rft_dat=xri:pqdiss:3319892.
Full textAbstract:
Thesis (Ph.D.)--Indiana University, Dept. of Physics, 2008.<br>Title from PDF t.p. (viewed on May 13, 2009). Source: Dissertation Abstracts International, Volume: 69-08, Section: B, page: 4823. Adviser: Shyh-Yuan Lee.
33
Gursoy, Ekrem Niebur Dagmar. "Independent component analysis for harmonic source identification in electric power systems /." Philadelphia, Pa. : Drexel University, 2007. http://hdl.handle.net/1860/1781.
Full text
34
Zhao, Yue. "Independent Component Analysis Enhancements for Source Separation in Immersive Audio Environments." UKnowledge, 2013. http://uknowledge.uky.edu/ece_etds/34.
Full textAbstract:
In immersive audio environments with distributed microphones, Independent Component Analysis (ICA) can be applied to uncover signals from a mixture of other signals and noise, such as in a cocktail party recording. ICA algorithms have been developed for instantaneous source mixtures and convolutional source mixtures. While ICA for instantaneous mixtures works when no delays exist between the signals in each mixture, distributed microphone recordings typically result various delays of the signals over the recorded channels. The convolutive ICA algorithm should account for delays; however, it requires many parameters to be set and often has stability issues. This thesis introduces the Channel Aligned FastICA (CAICA), which requires knowledge of the source distance to each microphone, but does not require knowledge of noise sources. Furthermore, the CAICA is combined with Time Frequency Masking (TFM), yielding even better SOI extraction even in low SNR environments. Simulations were conducted for ranking experiments tested the performance of three algorithms: Weighted Beamforming (WB), CAICA, CAICA with TFM. The Closest Microphone (CM) recording is used as a reference for all three. Statistical analyses on the results demonstrated superior performance for the CAICA with TFM. The algorithms were applied to experimental recordings to support the conclusions of the simulations. These techniques can be deployed in mobile platforms, used in surveillance for capturing human speech and potentially adapted to biomedical fields.
35
Li, Rui Fa. "Advanced process monitoring and control using principal and independent component analysis." Thesis, University of Leeds, 2003. http://ethos.bl.uk/OrderDetails.do?uin=uk.bl.ethos.275714.
Full text
36
Prasad, P. Shiva. "Independent Component Analysis." Thesis, 2007. http://ethesis.nitrkl.ac.in/52/2/10307008.pdf.
Full textAbstract:
A fundamental problem in neural network research, as well as in many other disciplines, is finding a suitable representation of multivariate data, i.e. random vectors. For reasons of computational and conceptual simplicity, the representation is often sought as a linear transformation of the original data. In other words, each component of the representation is a linear combination of the original variables. Well-known linear transformation methods include principal component analysis, factor analysis, and projection pursuit. Independent component analysis (ICA) is a recently developed method in which the goal is to find a linear representation of nongaussian data so that the components are statistically independent, or as independent as possible. Such a representation seems to capture the essential structure of the data in many applications,
including feature extraction and signal separation.
37
Huang, Huang-Wen, and 黃煌文. "Comparison between Network Component Analysis and Independent Component Analysis." Thesis, 2007. http://ndltd.ncl.edu.tw/handle/44160739111819490954.
Full textAbstract:
碩士<br>國立中正大學<br>統計科學所<br>95<br>Network component analysis (NCA) and independent component analysis (ICA) both are the ways of redundancy reduction. These statistical methods are for transforming an observed multidimensional random vector into lowerdimension. In this thesis, we use two different algorithms for linear ICA: fast fixed-point algorithm and joint approximate diagonalization of eigenmatrices algorithm. We compare these three techniques that the ability of
reconstructing the hidden regulatory layers, via the simulation studies and a real data examples. We also investigate the sensitivity of the reconstructed signals to inaccuracies in the strength of network connectivity.
38
chang, Lun-ching, and 張倫境. "Remarks on Network Component Analysis and Independent Component Analysis." Thesis, 2007. http://ndltd.ncl.edu.tw/handle/02695657099892855016.
Full textAbstract:
碩士<br>國立中正大學<br>統計科學所<br>96<br>In this thesis, two algorithms Fast ICA and Jade ICA, in independent component
analysis are performed to find the actual sources from the mixing
sources. It is well-known that the actual sources must be mutually independent
in ICA methods and be far away from the Gaussian distribution.
A new method called network component analysis (NCA) can also be applied
to blind sources cases without the mutually independence of sources
assumption. To illustrate the applications of these approaches, some simulation
studies and a real data example are provided in this thesis.
39
Pan, Jia-Chiun. "Covariate-Adjusted Independent Component Analysis." 2004. http://www.cetd.com.tw/ec/thesisdetail.aspx?etdun=U0001-0207200411343900.
Full text
40
Li, Shih-husiung, and 李仕雄. "Nonstationary Bayesian Independent Component Analysis." Thesis, 2009. http://ndltd.ncl.edu.tw/handle/15612008396397801403.
Full textAbstract:
碩士<br>國立成功大學<br>資訊工程學系碩博士班<br>97<br>In an intelligent speech perception system, it is required to recover speech signals from the mixed signals where some unknown and independent sources are simultaneously acquired by the system microphones. As we known, the independent component analysis (ICA) is a popular approach for blind source separation (BSS) and is referred as an important issue in the fields of machine learning. Traditionally, the standard ICA assumes that the source signals are stationary. This assumption restricts the performance of ICA in real-world applications. Since the source signals may be moving or may be active or inactive as time goes on, we propose a nonstationary Bayesian ICA (NB-ICA) for dealing with the nonstationary blind source separation for an intelligent speech perception system. The proposed NB-ICA algorithm is based on the online Bayesian learning theory which identifies the source activity and estimates the number of source signals in real time. Also, comparing with the other nonstationary ICA algorithms, the computation cost of the proposed NB-ICA can be significantly reduced. In addition, the segmentation information of the source signals is important for the applications of intelligent speech perception such as the systems of speech recognition, speaker identification and speaker diarization. Accordingly, we incorporate a semi-Markov model for capturing the duration information for the estimated status of the source signals. The experimental results show that the proposed NB-ICA algorithm is efficient for the separation of nonstationary source signals in terms of signal-to-inference ratio and detection accuracy of signal segments.
41
Pan, Jia-Chiun, and 潘家群. "Covariate-Adjusted Independent Component Analysis." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/92973510261499725567.
Full textAbstract:
碩士<br>國立臺灣大學<br>流行病學研究所<br>92<br>Independent component analysis (ICA) is a recently developed statistical and computational technique for discovering mutually independent nongaussian latent variables from observed multivariate data in the fields of neural networks and signal processing. It can potentially be applied to many application fields such as brain imaging, audio separation, telecommunication, feature extraction, economics, psychology, physiology, biomedical engineering, and bioinformatics, whenever the assumptions of statistical independence and nongaussianity are substantively justifiable. In current practice, one applies the standard procedure(s) of ICA directly to the observed multivariate variables, even though they may be affected by some known covariates, to identify the independent components and estimate the mixing coefficients. In this study, we find that ignoring those relevant covariates may lead to a biased result of ICA, and then suggest a covariate-adjusted ICA to minimize such biases by applying the standard procedure(s) of ICA to the residuals from the regressions of the observed multivariate variables on those relevant covariates in a linear ICA model. A simulation study is conducted using the FastICA algorithm to examine the statistical properties of our covariate-adjusted ICA and to derive numerically the sampling distributions of the estimated mixing coefficients as an interesting by-product. Finally, two examples are given for illustration.
42
Chia-Ho, Lin, and 林家合. "Independent Component Analysis and Its Applications." Thesis, 2003. http://ndltd.ncl.edu.tw/handle/18822129213916587385.
Full textAbstract:
碩士<br>國立臺灣大學<br>電信工程學研究所<br>91<br>When we want to find some relationship in large amounts of data, such as data analysis, feature extraction, signal processing, de-noising, and neural network research, a suitable representation of these data or data compression is needed. To achieve these goals, we usually do some linear transformation on original data. For example, principle component analysis, factor analysis, projection pursuit, and blind identification, etc. Recently, a powerful linear transform method called independent component analysis (ICA) is developed. We can say that this is an extension of principle component analysis (PCA), which can only impose uncorrelated up to second order statistics, and thus, defines orthogonal directions. However, ICA does not have these constrains and can analyze higher-order statistics.
Independent component analysis can be applied on many applications such as electroencephalogram (EEG), de-noising, remote sensing, watermarking (data encryption or decryption), feature extraction such as facial detection and recognition, texture extraction, and object classification, etc.
This thesis consists of five chapters. There will be an introduction to independent component analysis in chapter 1, including problem formulation, related works, some applications and the organization of this thesis. In chapter 2, I will introduce the algorithm of principle component analysis first, and then the theories and algorithms of independent component analysis will be explained. Also I will compare these algorithms and a summary will be made. In chapter 3, I will introduce applications of ICA on image processing, such as watermark, face detection, de-noise, etc. In chapter 4, I will introduce applications of ICA on sound processing, such as separating human speeches or separating human voice from background music .Finally, a concluding remark and future research directions shall be given in chapter 5.
43
Chen, Ching-Wen, and 陳清文. "Independent Component Analysis Applied to Meditation VEP Analysis." Thesis, 2004. http://ndltd.ncl.edu.tw/handle/8k8z2n.
Full text
44
Marcynuk, Kathryn L. "Independent component analysis for maternal-fetal electrocardiography." 2015. http://hdl.handle.net/1993/30181.
Full textAbstract:
Separating unknown signal mixtures into their constituent parts is a difficult problem in signal processing called blind source separation. One of the benchmark problems in this area is the extraction of the fetal heartbeat from an electrocardiogram in which it is overshadowed by a strong maternal heartbeat. This thesis presents a study of a signal separation technique called independent component analysis (ICA), in order to assess its suitability for the maternal-fetal ECG separation problem. This includes an analysis of ICA on deterministic, stochastic, simulated and recorded ECG signals. The experiments presented in this thesis demonstrate that ICA is effective on linear mixtures of known simulated or recorded ECGs. The performance of ICA was measured using visual comparison, heart rate extraction, and energy, information theoretic, and fractal-based measures. ICA extraction of clinically recorded maternal-fetal ECGs mixtures, in which the source signals were unknown, were successful at recovering the fetal heart rate.
45
"Extensions of independent component analysis: towards applications." Thesis, 2005. http://library.cuhk.edu.hk/record=b6074028.
Full textAbstract:
In practice, the application and extension of the ICA model depend on the problem and the data to be investigated. We finally focus on GARCH models in finance, and show that estimation of univariate or multivariate GARCH models is actually a nonlinear ICA problem; maximizing the likelihood is equivalent to minimizing the statistical dependence in standardized residuals. ICA can then be used for factor extraction in multivariate factor GARCH models. We also develop some extensions of ICA for this task. These techniques for extracting factors from multivariate return series are compared both theoretically and experimentally. We find that the one based on conditional decorrelation between factors behaves best.<br>In this thesis, first we consider the problem of source separation of post-nonlinear (PNL) mixtures, which is an extension of ICA to the nonlinear mixing case. With a large number of parameters, existing methods are computation-demanding and may be prone to local optima. Based on the fact that linear mixtures of independent variables tend to be Gaussian, we develop a simple and efficient method for this problem, namely extended Gaussianization. With Gaussianization as preprocessing, this method approximates each linear mixture of independent sources by the Cornish-Fisher expansion with only two parameters. Inspired by the relationship between the PNL mixing model and the Wiener system, extended Gaussianization is also proposed for blind inversion of Wiener systems.<br>Independent component analysis (ICA) is a recent and powerful technique for recovering latent independent sources given only their mixtures. The basic ICA model assumes that sources are linearly mixed and mutually independent.<br>Next, we study the subband decomposition ICA (SDICA) model, which extends the basic ICA model to allow dependence between sources by assuming that only some narrow-band source sub-components are independent. In SDICA, it is difficult to determine the subbands of source independent sub-components. We discuss the feasibility of performing SDICA in an adaptive manner. An adaptive method, called band selective ICA, is then proposed for this task. We also investigate the relationship between overcomplete ICA and SDICA and show that band selective ICA can solve the overcomplete ICA problems with sources having specific frequency localizations. Experimental results on separating images of human faces as well as artificial data are presented to verify the powerfulness of band selective ICA.<br>Zhang Kun.<br>"July 2005."<br>Adviser: Lai-Wan Chan.<br>Source: Dissertation Abstracts International, Volume: 67-07, Section: B, page: 3925.<br>Thesis (Ph.D.)--Chinese University of Hong Kong, 2005.<br>Includes bibliographical references (p. 218-234).<br>Electronic reproduction. Hong Kong : Chinese University of Hong Kong, [2012] System requirements: Adobe Acrobat Reader. Available via World Wide Web.<br>Electronic reproduction. [Ann Arbor, MI] : ProQuest Information and Learning, [200-] System requirements: Adobe Acrobat Reader. Available via World Wide Web.<br>Abstract in English and Chinese.<br>School code: 1307.
46
Yang, Yi-Cyun, and 楊逸群. "Nonlinear Independent Component Analysis using Generalized Adalines." Thesis, 2006. http://ndltd.ncl.edu.tw/handle/37318403517918691791.
Full textAbstract:
碩士<br>國立東華大學<br>應用數學系<br>94<br>A new method is devised for linear and post-nonlinear independent component analysis. Unlike traditional statistics oriented ICA algorithms, which have been developed based on minimization of the Kullback-Leibler(KL) divergence between retrieved components, this method uses the recurrent optimal post-nonlinear kernel method to realize blind separation of linear or post-nonlinear mixtures of independent sources. The post-nonlinear mixing structure of independent sources is realized by multiple generalized adalines(gadalines). Following the leave-one-out learning strategy, hyper-parameters of each gadaline as well as intermediate independent components are optimized under a mean-field-annealing process to satisfy constraints given by multi-channel observations. The new method is shown more feasible for post-nonlinear independent component analysis. It indeed involves with no computations of optimizing statistics, such as the KL divergence or high order moments, which have been known diffcult to be resolved for nonlinear independent component analysis.
47
Ejaz, Masood. "A framework for implementing Independent Component Analysis algorithms." 2008. http://etd.lib.fsu.edu/theses/available/etd-04182008-123550.
Full textAbstract:
Thesis (Ph. D.)--Florida State University, 2008.<br>Advisor: Simon y. Foo, Florida State University, FAMU-FSU College of Engineering, Dept. of Electrical & Computer Engineering. Title and description from dissertation home page (viewed July 23, 2008). Document formatted into pages; contains viii,120 pages. Includes bibliographical references.
48
Kaplan, Sam T. "Principal and independent component analysis for seismic data." Thesis, 2003. http://hdl.handle.net/2429/14100.
Full textAbstract:
Principal and Independent component analysis (PCA and ICA) are two ideas which are very much related; both employing a statistical understanding of data to achieve their goals. Whereas PCA exploits statistical correlation, ICA uses statistical independence to glean useful information from data. Seismic data is inherently noisy, and is complicated by the presence of an unknown seismic wavelet. Analysis of the data is aided by, both, noise suppression and blind deconvolution techniques. First, consider the subject of noise suppression. If the data are organized into several sequences where, from one sequence to the next, the signal is correlated while the noise is uncorrelated, then PCA has the ability to separate noise and signal. Here, PCA is analyzed from three points of view, variance maximization, the singular value decomposition and neural networks. The resulting theory is used to filter noise from a set of common midpoint seismic gathers by exploiting correlations which exist from one gather to the next. To further simplify analysis of these data, the Earth is often approximated as a linear system; thus, the seismic trace is subject to the convolutional model. Convolution is a linear operation, and consequently, can be formulated as a linear system of equations. If only the output of the system (the convolved signal) is known, then the problem is blind so that given one equation, two unknowns are sought. This problem is well suited for ICA which has the ability to find some estimate of the two unknowns, and here the blind deconvolution problem is solved using ICA. To facilitate this, several time-lagged versions of the convolved signal are extracted and used to construct realizations of a random vector. For ICA, this random vector is the, so called, mixture vector, created by the matrix-vector multiplication of the two unknowns, the mixing matrix and the source vector. Due to the properties of convolution, the mixing matrix is banded with its nonzero elements containing the convolution's filter. This banded property is incorporated into the ICA algorithm as prior information, giving rise to a banded ICA algorithm (B-ICA) which is, in turn, used in a new blind deconvolution algorithm. This algorithm is considered for both noiseless and noisy data.
49
Lee, Shih-Hua, and 李世驊. "A new algorithm for convolutive independent component analysis." Thesis, 2006. http://ndltd.ncl.edu.tw/handle/42596436345836454563.
Full textAbstract:
碩士<br>國立東華大學<br>應用數學系<br>94<br>This work addresses on blind separation of convolutive mixtures of independent sources. The temporally convolutive structure is assumed to be composed of multiple mixing matrices, each corresponding to a time delay, collectively transforming a segment of consecutive source signals to form multi-channel observations. As τ=1, this problem reduces to linear independent component analysis. For arbitrary τ, we propose a new algorithm to estimate the unknown convolutive structure as well as independent sources. The proposed convolutive ICA algorithm is based on optimal kernel estimation and leave-one-out approximation operated under the mean-field-annealing process. We test the new algorithm with artificially created data and two-microphone recordings of speech and musics. It is shown that the error between the estimated and given convolutive structures is significantly reduced for artificially created data and the human speech is well separated from the background musics for the two-microphone recordings. The proposed new algorithm is empirically shown effective for blind separation of convolutive mixtures of independent sources.
Keyword: Convolutive mixture, independent component analysis, blind separation, optimal kernel estimation, leave-one-out approximation, K-state transfer function, mean field annealing.
50
Chien, Wei-Nan, and 簡瑋男. "Chinese Near-Synonym Substitution Using Independent Component Analysis." Thesis, 2011. http://ndltd.ncl.edu.tw/handle/87414927958735528785.
Full textAbstract:
碩士<br>元智大學<br>資訊管理學系<br>99<br>Near-synonym sets represent groups of words with similar meaning, which are useful knowledge resources for many natural language applications such as query expansion for information retrieval (IR) and computer-assisted language learning. However, near-synonyms are not necessarily interchangeable in contexts due to their specific usage and syntactic constraints. Previous studies have developed various methods for near-synonym choice in English sentences. To our best knowledge, there is no such evaluation on Chinese sentences. Therefore, this paper proposes the use of the independent component analysis (ICA) for Chinese near-synonym choice evaluation. Experimental results show that the ICA achieves higher accuracy than the pointwise mutual information (PMI), 5-gram language model and vector space model (VSM) that have been used in previous studies.