FORSCHUNGSZENTRUM JÜLICH GmbH Zentralinstitut für Angewandte Mathematik D Jülich, Tel. (02461)
|
|
- Amber Watts
- 5 years ago
- Views:
Transcription
1 FORSCHUNGSZENTRUM JÜLICH GmbH Zentralinstitut für Angewandte Mathematik D Jülich, Tel. (2461) Interner Bericht Temporal and Spatial Prewhitening of Multi-Channel MEG Data Roland Beucker, Heidi Anne Schlitt* FZJ-ZAM-IB-9721 November 1997 (letzte Änderung: ) (*) Picker International, MR Division, 595 Miner Rd., Highland Heights, OH 44143, USA
2
3 Temporal and Spatial Prewhitening of Multi-Channel MEG Data R. Beucker and H. A. Schlitt y Central Institute for Applied Mathematics Research Centre Julich GmbH D Julich, Germany y Picker International MR Division 595 Miner Rd. Highland Heights, OH 44143, USA Abstract In this paper we present a prewhitening strategy to decorrelate the noise in space and in time. We discuss in detail how the eects of the ltering can be seen in the singular value decomposition of the data and we demonstrate how the approach can be used with an objective, statistical signal subspace estimation algorithm. 1 Introduction For multiple dipole reconstruction methods the noise is assumed to be spatially and temporally white. The statistical model for Magnetoencephalography (MEG) data introduced by Mosher et al. [1] consists of time series to which spatially and temporally zero-mean white noise is added. Under ideal conditions, in the Value Decomposition (SVD) spectrum of measured MEG data a singular value with a high multiplicity represents the noise part of the spectrum. For real measured data, however, the noise is not white and the noise singular values do not have identical values. By prewhitening the data, the noise can be made to more closely meet the assumption of being spatially and temporally white. Mosher et al. [2], [1] proposed, but did not implement, prewhitening the measured MEG data before performing a reconstruction with Multiple Signal Characterization 1
4 (MUSIC). In order to use MUSIC, one has to predict the dimension of the signal subspace [2]. The dimension can be determined by visual inspection of the singular value spectrum or by applying a statistical test. We found that, in contrast to the 'Minimum Description Length' (MDL), `Akaike Information Criterion' (AIC) [3] or several other tests (e.g. sphericity test), the `Fast Subspace Decomposition' (FSD) from Xu et al. [4] produces good results [5], [6]. To improve the statistical condition for the application of the test and for the MUSIC reconstruction we present a strategy for temporal and spatial prewhitening of multi-channel MEG-data [7]. We demonstrate the eects of prewhitening on two sets of measured data. 2 Methods We use the time series model X t = m t + N t introduced by Mosher et al. [1] which consists of a deterministic trend component, m t, dependent on the time courses of the associated sources and a random noise component, N t. To temporally prewhiten the data we subtract the deterministic trend, model the noise with a vector autoregressive process (VAR), determine the model order with the 'bias corrected Akaike information criterion' (AICC), estimate the noise covariance matrix and apply the VAR model as a lter to the measured data. We determine the coecient matrices of the VAR-process by taking the multi-dimensional Yule-Walker estimator [8]. The spatial prewhitening consists of an orthogonality transformation and the normalization of the noise eigenvalues [8]. Prewhitening strategy: 1. Estimation and elimination of the deterministic trend m t Filter the measured time slices with a low pass lter Subtract the ltered data from the measured data to form an estimate of the noise ( ^N t ) 2. Model the noise by a vector autoregressive process (VAR) of order p N t = P p j=1 jn t?j + w t Determine the order, p, with the bias corrected Akaike information criterion (AICC) and the coecient matrices, j, using ^Nt Estimate covariance matrix of the white noise w t 3. Temporal prewhitening: Application of the VAR-Model as lter Filter the measured data with the help of the determined coecient matrices 1; : : : ; p 4. Spatial prewhitening: Diagonalization of 2
5 Calculate the eigendecomposition of = V D V T Apply D? 1 2 V T to the temporally prewhitened measured data It is common to estimate the noise covariance matrix from data measured before stimulus onset or to measure the environment noise for several hundreds of milliseconds without a subject [9]. Because the noise conditions change with time, we decided to subtract the deterministic trend of the time interval which contains the evoked response. If one assumes the noise is not strongly correlated over time, one can directly calculate the noise covariance matrix and multiply the data with D? 1 2 V T (spatial prewhitening) [9]. This has the eect that the singular values which one assumes to belong to the noise are normalized (gure 2). The spatial prewhitening also eliminates the spatial dependencies of the sensors between each other (orthogonality transformation). The temporal prewhitening eliminates dependencies in the noise in time and gives the researcher an indication which singular values represented dependencies in the noise after subtracting the global trend. 3 Results We rst analyzed data from an auditory experiment to show the eect of spatial prewhitening on the SVD. These data were measured with the Neuromag 122 channel whole-head array [1], but we restricted our analysis to one of the two sets of orthogonal gradient directions (61 channels) to reduce the spatial dependence of the noise. The measured excitation was evoked by a 5 ms 1 khz sinusoidal sound (left ear). The measurement was started 1 ms before stimulus onset. The sampling frequency was 297 Hz and a band-pass lter (.3-9 Hz) was applied to the data. For our further analysis we averaged 95 epochs. One epoch lasted 6 ms (18 time slices, gure 1). The SVD of the auditory data (gure 2) shows that the small singular values continue to decrease instead of leveling out as they would if the noise were white. Before applying the FSD to these data we decorrelate the noise in space. As an estimate for the noise we simply took the residuals after subtracting the deterministic trend from the measured data and calculated the noise covariance matrix. After spatial prewhitening, however, the small singular values have approximately the same size. The signal subspace dimension estimated with the FSD is 4. Before the spatial prewhitening the estimated signal subspace dimension was 5. Because of the fact that our chosen interval contains a long pre- and poststimulus intervals, background activity is also represented by singular values which can be seen as signals. If the interval size is reduced to include only the evoked response, the estimated signal subspace dimension becomes smaller. Temporally prewhitening of the data was 3
6 8 6 A m 4 p l 2 i t u d -2 e -4 f T / cm ms f T 4 3 A 2 m p 1 l i t -1 u d-2 e ms Figure 1: On the left side data from an auditory experiment is presented. 6ms with a sampling rate from 297Hz are presented. On the right 768ms of averaged data from a somato-sensory experiment are presented not possible because the sampling rate was too low to estimate parameters for the autoregressive model. 7 1 of Values of Values Noise Part estimated by the FSD Figure 2: On the left side the SVD of 61 channels of the auditory data is shown. The smaller singular values fall o strongly. On the right the eect of spatial prewhitening can be seen. Because of the normalization the singular values have a dierent scaling. The smaller singular values of the whole spectrum have approximately the same size. The second data set was from a somato-sensory experiment measured with a BTi 37 channel array. The measured excitation was evoked by stimulating the thumb with pus of air. The epochs were 2 s long, the measurement was started 1 ms before stimulus onset, and the response occurred at about 13 ms post-stimulus. The sampling frequency was 52 Hz and 141 time slices were recorded per epoch. During the measurement, a high pass lter with a cut-o frequency of 1 Hz was applied. We averaged 2 epochs and chose a subinterval from 1 ms to 768 ms (4 time slices) for our further analysis. This subinterval is shown in gure 1. 4
7 7 6 5 of 4 Values Figure 3: In this gure the complete SVD spectrum of the non-ltered somatosensory data is shown of Values 4 Temporally Prewhitened of Values Temporally Prewhitened Spatially Prewhitened Figure 4: In this gure on the left side the SVD spectrum of the temporally prewhitened time series from the somato-sensory experiment is shown. On the right side the spectrum of the temporally and spatially prewhitened time series is shown. In gure 3 the SVD of the somato-sensory data is shown. One cannot clearly see which singular values belong to the signal part and which to the noise part. In gure 4 we plotted the singular values from the temporally prewhitened data and those ones from the temporally and spatially prewhitened data. After applying the rst step of our algorithm (temporal prewhitening, model order p = 2) the distance from the sixth to the seventh singular value of the temporally prewhitened data decreases. This indicates that the noise residuals resulting after subtracting the global trend from the data were not independent. The decorrelation in time separates the spectrum which makes it much easier to decide which singular values belong to which part. After spatial prewhitening the normalization eect can clearly be seen. The estimated signal subspace dimension of the chosen interval of the somatosensory data was 15. After temporal and spatial prewhitening the estimated signal subspace dimension of the chosen interval was 14. 5
8 4 Discussion In this paper we presented a method for temporal and spatial prewhitening. After subtracting the global trend we showed how to model the noise by a vector autoregressive process and how the spectrum is changed after applying the VAR-process as a lter to the measured data. We showed that in case of lower temporal sample rates the spatial prewhitening can be used to equalize the small singular values of the spectrum. Our prewhitening strategy improves the statistical conditions so that statistical methods for estimating the dimension of the signal subspace (e.g. FSD) and multiple dipole reconstruction methods (e.g. MUSIC) can produce better results. The spatial prewhitening normalizes the variances and makes the small singular values more equal. The noise covariance matrix, which must be known for the spatial prewhitening, can be determined with the help of the VAR-model. If the data has been sampled at a high enough rate in time, then temporal prewhitening can be used to improve the noise statistics. If there are not enough time samples, the coecient matrices cannot be estimated in a stable fashion. In this case the noise covariance matrix can be estimated either by collecting a prestimulus interval [9] or by subtracting the global trend of the time interval which contains the evoked response or by using a recursive least squares algorithm [11]. We recommend the prewhitening before applying multiple dipole methods because, in general, the measurement noise is correlated. Furthermore, one should use a sampling rate from 512 Hz so that one can use approximately 4 time slices for 37 channels for the calculation of the coecient matrices of the VAR-model. 5 Acknowledgments The authors are indebted to W. Meyer (Central Institute for Applied Mathematics, Research Centre Julich) and J. C. Mosher (Biophysics Group, Los Alamos National Laboratory) for giving many valuable comments. We would like to thank R. Hari (Low Temperature Laboratory, Helsinki University of Technology) and K. Wassmuth (Biomagnetic Technologies, Aachen) for generously sharing MEG data with us, and F. Hofeld (Central Institute for Applied Mathematics, Research Centre Julich) for his support. References [1] J. C. Mosher and R. M. Leahy. EEG and MEG source localization using recursively applied (RAP) MUSIC. Proceedings of the Thirtieth Annual Asilomar Conference on Signals, Systems, and Computers, Pacic Grove, California, USA, November
9 [2] J. C. Mosher, P. S. Lewis, and R. M. Leahy. Multiple dipole modeling and localization from spatio-temporal MEG data. IEEE Trans. Biomed. Eng., 39:541{ 557, [3] M. Wax and T. Kailath. Detection of signals by information theoretic criteria. IEEE Trans. Acoust. Speech Signal Process., 33:387{392, [4] G. Xu and T. Kailath. Fast estimation of principal eigenspace using Lanczos algorithm. SIAM J. Matrix Anal. Appl., 15:974{994, [5] R. Beucker and H. A. Schlitt. Objective signal subspace determination for meg. Technical Report FZJ-ZAM-IB-9715, Research Centre Julich, Julich, Germany, [6] R. Beucker and H. A. Schlitt. Comparison of statistical tests for determining the number of MEG sources. In Abstract Book, Eighth World Congress of the International Society for Brain Electromagnetic Topography, Zurich, Switzerland, March [7] R. Beucker and H. A. Schlitt. Spatial and temporal prewhitening of multichannel EEG/MEG data. In Abstract Book, Sixth German EEG/EP Mapping Meeting, Gieen, Germany, September [8] P. J. Brockwell and R. A. Davis. Time Series: Theory and Methods. Springer, second edition, [9] K. Sekihara, D. Poeppel, A. Marantz, H. Koizumi, and Y. Miyashita. Noise covariance incorporated MEG-MUSIC algorithm: A method for multiple-dipole estimation tolerant of the inuence of background brain activity. IEEE Trans. Biomed. Eng., 44:839{847, [1] M. S. Hamalainen. Functional localization based on measurement with a wholehead magnetometer system. Brain Topogr., 7:283{289, [11] M. Campi. Performance of RLS identication algorithms with forgetting factor: A -mixing approach. J. Math. Systems Estim. Control, 7:29{53,
Paired MEG Data Set Source Localization Using Recursively Applied and Projected (RAP) MUSIC
1248 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 47, NO. 9, SEPTEMBER 2000 Paired MEG Data Set Source Localization Using Recursively Applied and Projected (RAP) MUSIC John J. Ermer, Member, IEEE,
More informationMaximum-Likelihood Estimation of Low-Rank Signals for Multiepoch MEG/EEG Analysis
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 51, NO. 11, NOVEMBER 2004 1981 Maximum-Likelihood Estimation of Low-Rank Signals for Multiepoch MEG/EEG Analysis Boris V. Baryshnikov*, Barry D. Van Veen,
More informationALTHOUGH the single equivalent current dipole (ECD)
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL 44, NO 9, SEPTEMBER 1997 839 Noise Covariance Incorporated MEG-MUSIC Algorithm: A Method for Multiple-Dipole Estimation Tolerant of the Influence of Background
More informationTHE PROBLEM of localizing the sources of event related. Recursive MUSIC: A Framework for EEG and MEG Source Localization
1342 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 45, NO. 11, NOVEMBER 1998 Recursive MUSIC: A Framework for EEG and MEG Source Localization John C. Mosher,* Member, IEEE, and Richard M. Leahy, Member,
More informationRecent advances in the analysis of biomagnetic signals
Recent advances in the analysis of biomagnetic signals Kensuke Sekihara Mind Articulation Project, Japan Science and echnology Corporation okyo Metropolitan Institute of echnology Application of spatial
More information(a)
Chapter 8 Subspace Methods 8. Introduction Principal Component Analysis (PCA) is applied to the analysis of time series data. In this context we discuss measures of complexity and subspace methods for
More information840 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 5, MAY 2006
840 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 53, NO. 5, MAY 2006 Performance Analysis of Reduced-Rank Beamformers for Estimating Dipole Source Signals Using EEG/MEG David Gutiérrez, Member, IEEE,
More informationPARTICLE FILTERING, BEAMFORMING AND MULTIPLE SIGNAL CLASSIFICATION FOR THE ANALYSIS OF MAGNETOENCEPHALOGRAPHY TIME SERIES: A COMPARISON OF ALGORITHMS
Volume X, No. X, X, X XX Web site: http://www.aimsciences.org PARTICLE FILTERING, BEAMFORMING AND MULTIPLE SIGNAL CLASSIFICATION FOR THE ANALYSIS OF MAGNETOENCEPHALOGRAPHY TIME SERIES: A COMPARISON OF
More informationSource Localization Using Recursively Applied and Projected (RAP) MUSIC
Source Localization Using Recursively Applied and Projected (RAP) MUSIC John C. Masher* and Richard M. Leahy *Los Alamos National Laboratory, Group P-1 MS D454, Los Alamos, NM 87545 Signal & Image Processing
More informationSource Localization Using Recursively Applied and Projected (RAP) MUSIC
Source Localization Using Recursively Applied and Projected (RAP) MUSIC John C. Mosher * and Richard M. Leahy + * Los Alamos National Laboratory Group P-21 MS D454 Los Alamos, NM 87545 mosher@lanl.gov,
More informationA Rao-Blackwellized particle filter for magnetoencephalography
A Rao-Blackwellized particle filter for magnetoencephalography C Campi Dipartimento di Matematica, Università di Genova, via Dodecaneso 35, 16146 Genova, Italy A Pascarella Dipartimento di Matematica,
More informationEstimating Evoked Dipole Responses in Unknown Spatially Correlated Noise with EEG/MEG Arrays
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 48, NO. 1, JANUARY 2000 13 Estimating Evoked Dipole Responses in Unknown Spatially Correlated Noise with EEG/MEG Arrays Aleksar Dogžić, Student Member, IEEE,
More informationBEAMFORMING DETECTORS WITH SUBSPACE SIDE INFORMATION. Andrew Bolstad, Barry Van Veen, Rob Nowak
BEAMFORMING DETECTORS WITH SUBSPACE SIDE INFORMATION Andrew Bolstad, Barry Van Veen, Rob Nowak University of Wisconsin - Madison 45 Engineering Drive Madison, WI 5376-69 akbolstad@wisc.edu, vanveen@engr.wisc.edu,
More informationPOLYNOMIAL SINGULAR VALUES FOR NUMBER OF WIDEBAND SOURCES ESTIMATION AND PRINCIPAL COMPONENT ANALYSIS
POLYNOMIAL SINGULAR VALUES FOR NUMBER OF WIDEBAND SOURCES ESTIMATION AND PRINCIPAL COMPONENT ANALYSIS Russell H. Lambert RF and Advanced Mixed Signal Unit Broadcom Pasadena, CA USA russ@broadcom.com Marcel
More informationBeamforming Techniques Applied in EEG Source Analysis
Beamforming Techniques Applied in EEG Source Analysis G. Van Hoey 1,, R. Van de Walle 1, B. Vanrumste 1,, M. D Havé,I.Lemahieu 1 and P. Boon 1 Department of Electronics and Information Systems, University
More informationSpatial Harmonic Analysis of EEG Data
Spatial Harmonic Analysis of EEG Data Uwe Graichen Institute of Biomedical Engineering and Informatics Ilmenau University of Technology Singapore, 8/11/2012 Outline 1 Motivation 2 Introduction 3 Material
More informationMultiple Signal Classification Algorithm Based Electric Dipole Source Localization Method in an Underwater Environment
S S symmetry Article Multiple Signal Classification Algorithm Based Electric Dipole Source Localization Method in an Underwater Environment Yidong Xu 1, Wei Xue 1, *, Yingsong Li 1,2, *, Lili Guo 1 and
More informationConfidence Interval of Single Dipole Locations Based on EEG Data
Brain Topography, Volume 10, Number 1,1997 31 Confidence Interval of Single Dipole Locations Based on EEG Data Christoph Braun*, Stefan Kaiser*, WilhelmEmil Kineses*, and Thomas Elbert^ Summary: Noise
More informationCONTENTS NOTATIONAL CONVENTIONS GLOSSARY OF KEY SYMBOLS 1 INTRODUCTION 1
DIGITAL SPECTRAL ANALYSIS WITH APPLICATIONS S.LAWRENCE MARPLE, JR. SUMMARY This new book provides a broad perspective of spectral estimation techniques and their implementation. It concerned with spectral
More informationUNIVERSAL FOURTH ORDER MUSIC METHOD : INCORPORATION OF ICA INTO MEG INVERSE SOLUTION. Satoshi Niijima, Shoogo Ueno
UNIVERSAL OURTH ORDER MUSIC METHOD : INCORPORATION O ICA INTO MEG INVERSE SOLUTION Satoshi Niijima, Shoogo Ueno Department of Electrical Engineering, Graduate School of Engineering, University of Tokyo
More informationTransformation of Whole-Head MEG Recordings Between Different Sensor Positions
Transformation of Whole-Head MEG Recordings Between Different Sensor Positions Transformation von Ganzkopf-MEG-Messungen zwischen verschiedenen Sensorpositionen Thomas R. Knösche Max Planck Institute of
More informationAcoustic MIMO Signal Processing
Yiteng Huang Jacob Benesty Jingdong Chen Acoustic MIMO Signal Processing With 71 Figures Ö Springer Contents 1 Introduction 1 1.1 Acoustic MIMO Signal Processing 1 1.2 Organization of the Book 4 Part I
More informationTHE PROBLEMS OF ROBUST LPC PARAMETRIZATION FOR. Petr Pollak & Pavel Sovka. Czech Technical University of Prague
THE PROBLEMS OF ROBUST LPC PARAMETRIZATION FOR SPEECH CODING Petr Polla & Pavel Sova Czech Technical University of Prague CVUT FEL K, 66 7 Praha 6, Czech Republic E-mail: polla@noel.feld.cvut.cz Abstract
More informationTO describe the underlying processes of electroencephalogram
414 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 3, MARCH 2005 Model Selection in Spatio-Temporal Electromagnetic Source Analysis Lourens J. Waldorp*, Member, IEEE, Hilde M. Huizenga, Arye
More informationEstimating Neural Sources from Each Time-Frequency Component of Magnetoencephalographic Data
642 IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 47, NO. 5, MAY 2000 Estimating Neural Sources from Each Time-Frequency Component of Magnetoencephalographic Data Kensuke Sekihara*, Member, IEEE, Srikantan
More informationIEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 3, MARCH
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 52, NO. 3, MARCH 2005 471 Distinguishing Between Moving and Stationary Sources Using EEG/MEG Measurements With an Application to Epilepsy İmam Şamil Yetik*,
More informationTutorial on Blind Source Separation and Independent Component Analysis
Tutorial on Blind Source Separation and Independent Component Analysis Lucas Parra Adaptive Image & Signal Processing Group Sarnoff Corporation February 09, 2002 Linear Mixtures... problem statement...
More informationNew Machine Learning Methods for Neuroimaging
New Machine Learning Methods for Neuroimaging Gatsby Computational Neuroscience Unit University College London, UK Dept of Computer Science University of Helsinki, Finland Outline Resting-state networks
More informationMEG Source Localization Using an MLP With Distributed Output Representation
MEG Source Localization Using an MLP With Distributed Output Representation Sung Chan Jun, Barak A. Pearlmutter, Guido Nolte CBLL Meeting October 5, 2005 Source localization Definition: Identification
More informationIndependent Component Analysis. Contents
Contents Preface xvii 1 Introduction 1 1.1 Linear representation of multivariate data 1 1.1.1 The general statistical setting 1 1.1.2 Dimension reduction methods 2 1.1.3 Independence as a guiding principle
More informationA Subspace Approach to Estimation of. Measurements 1. Carlos E. Davila. Electrical Engineering Department, Southern Methodist University
EDICS category SP 1 A Subspace Approach to Estimation of Autoregressive Parameters From Noisy Measurements 1 Carlos E Davila Electrical Engineering Department, Southern Methodist University Dallas, Texas
More informationPrincipal Component Analysis. Applied Multivariate Statistics Spring 2012
Principal Component Analysis Applied Multivariate Statistics Spring 2012 Overview Intuition Four definitions Practical examples Mathematical example Case study 2 PCA: Goals Goal 1: Dimension reduction
More informationOn Moving Average Parameter Estimation
On Moving Average Parameter Estimation Niclas Sandgren and Petre Stoica Contact information: niclas.sandgren@it.uu.se, tel: +46 8 473392 Abstract Estimation of the autoregressive moving average (ARMA)
More informationOn the Behavior of Information Theoretic Criteria for Model Order Selection
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 49, NO. 8, AUGUST 2001 1689 On the Behavior of Information Theoretic Criteria for Model Order Selection Athanasios P. Liavas, Member, IEEE, and Phillip A. Regalia,
More informationADAPTIVE FILTER THEORY
ADAPTIVE FILTER THEORY Fourth Edition Simon Haykin Communications Research Laboratory McMaster University Hamilton, Ontario, Canada Front ice Hall PRENTICE HALL Upper Saddle River, New Jersey 07458 Preface
More informationIEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 3, MARCH
IEEE TRANSACTIONS ON BIOMEDICAL ENGINEERING, VOL. 55, NO. 3, MARCH 2008 1103 Array Response Kernels for EEG and MEG in Multilayer Ellipsoidal Geometry David Gutiérrez*, Member, IEEE, and Arye Nehorai,
More informationIn: W. von der Linden, V. Dose, R. Fischer and R. Preuss (eds.), Maximum Entropy and Bayesian Methods, Munich 1998, Dordrecht. Kluwer, pp
In: W. von der Linden, V. Dose, R. Fischer and R. Preuss (eds.), Maximum Entropy and Bayesian Methods, Munich 1998, Dordrecht. Kluwer, pp. 17-6. CONVERGENT BAYESIAN FORMULATIONS OF BLIND SOURCE SEPARATION
More information/16/$ IEEE 1728
Extension of the Semi-Algebraic Framework for Approximate CP Decompositions via Simultaneous Matrix Diagonalization to the Efficient Calculation of Coupled CP Decompositions Kristina Naskovska and Martin
More informationRealtime MEG source localization with realistic noise
Realtime MEG source localization with realistic noise Sung Chan Jun Barak A. Pearlmutter Guido Nolte Department of Computer Science University of New Mexico Albuquerque, NM 87 junsc,bap,nolte @cs.unm.edu
More informationFAST AND EFFECTIVE MODEL ORDER SELECTION METHOD TO DETERMINE THE NUMBER OF SOURCES IN A LINEAR TRANSFORMATION MODEL
FAST AND EFFECTIVE MODEL ORDER SELECTION METHOD TO DETERMINE THE NUMBER OF SOURCES IN A LINEAR TRANSFORMATION MODEL Fengyu Cong 1, Asoke K Nandi 1,2, Zhaoshui He 3, Andrzej Cichocki 4, Tapani Ristaniemi
More informationThe Mathematics of Facial Recognition
William Dean Gowin Graduate Student Appalachian State University July 26, 2007 Outline EigenFaces Deconstruct a known face into an N-dimensional facespace where N is the number of faces in our data set.
More informationPrincipal Component Analysis
Principal Component Analysis Yingyu Liang yliang@cs.wisc.edu Computer Sciences Department University of Wisconsin, Madison [based on slides from Nina Balcan] slide 1 Goals for the lecture you should understand
More informationA METHOD OF ADAPTATION BETWEEN STEEPEST- DESCENT AND NEWTON S ALGORITHM FOR MULTI- CHANNEL ACTIVE CONTROL OF TONAL NOISE AND VIBRATION
A METHOD OF ADAPTATION BETWEEN STEEPEST- DESCENT AND NEWTON S ALGORITHM FOR MULTI- CHANNEL ACTIVE CONTROL OF TONAL NOISE AND VIBRATION Jordan Cheer and Stephen Daley Institute of Sound and Vibration Research,
More informationTime Series: Theory and Methods
Peter J. Brockwell Richard A. Davis Time Series: Theory and Methods Second Edition With 124 Illustrations Springer Contents Preface to the Second Edition Preface to the First Edition vn ix CHAPTER 1 Stationary
More informationPCA, Kernel PCA, ICA
PCA, Kernel PCA, ICA Learning Representations. Dimensionality Reduction. Maria-Florina Balcan 04/08/2015 Big & High-Dimensional Data High-Dimensions = Lot of Features Document classification Features per
More informationComparative Performance Analysis of Three Algorithms for Principal Component Analysis
84 R. LANDQVIST, A. MOHAMMED, COMPARATIVE PERFORMANCE ANALYSIS OF THR ALGORITHMS Comparative Performance Analysis of Three Algorithms for Principal Component Analysis Ronnie LANDQVIST, Abbas MOHAMMED Dept.
More informationA New High-Resolution and Stable MV-SVD Algorithm for Coherent Signals Detection
Progress In Electromagnetics Research M, Vol. 35, 163 171, 2014 A New High-Resolution and Stable MV-SVD Algorithm for Coherent Signals Detection Basma Eldosouky, Amr H. Hussein *, and Salah Khamis Abstract
More informationADAPTIVE FILTER THEORY
ADAPTIVE FILTER THEORY Fifth Edition Simon Haykin Communications Research Laboratory McMaster University Hamilton, Ontario, Canada International Edition contributions by Telagarapu Prabhakar Department
More informationBLOCK-BASED MULTICHANNEL TRANSFORM-DOMAIN ADAPTIVE FILTERING
BLOCK-BASED MULTICHANNEL TRANSFORM-DOMAIN ADAPTIVE FILTERING Sascha Spors, Herbert Buchner, and Karim Helwani Deutsche Telekom Laboratories, Technische Universität Berlin, Ernst-Reuter-Platz 7, 10587 Berlin,
More informationDetection in reverberation using space time adaptive prewhiteners
Detection in reverberation using space time adaptive prewhiteners Wei Li,,2 Xiaochuan Ma, Yun Zhu, Jun Yang,,2 and Chaohuan Hou Institute of Acoustics, Chinese Academy of Sciences 2 Graduate University
More informationHIGH-SPEED data transmission and real-time multimedia
5658 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 55, NO 12, DECEMBER 2007 Reduced-Rank MDL Method for Source Enumeration in High-Resolution Array Processing Lei Huang, Member, IEEE, Shunjun Wu, Member,
More informationUMIACS-TR July CS-TR 2494 Revised January An Updating Algorithm for. Subspace Tracking. G. W. Stewart. abstract
UMIACS-TR-9-86 July 199 CS-TR 2494 Revised January 1991 An Updating Algorithm for Subspace Tracking G. W. Stewart abstract In certain signal processing applications it is required to compute the null space
More informationSENSITIVITY COMPUTATIONS FOR ELLIPTIC EQUATIONS WITH INTERFACES. Virginia Polytechnic Institute and State University Blacksburg, VA,
SENSITIVITY COMPUTATIONS FOR ELLIPTIC EQUATIONS WITH INTERFACES J.A. Burns 1, D. Rubio 2 and M. I. Troparevsky 3 1 Interdisciplinary Center for Applied Mathematics Virginia Polytechnic Institute and State
More informationEfficient and Accurate Rectangular Window Subspace Tracking
Efficient and Accurate Rectangular Window Subspace Tracking Timothy M. Toolan and Donald W. Tufts Dept. of Electrical Engineering, University of Rhode Island, Kingston, RI 88 USA toolan@ele.uri.edu, tufts@ele.uri.edu
More informationEEG/MEG Error Bounds for a Static Dipole Source with a Realistic Head Model
470 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 49, NO. 3, MARCH 2001 EEG/MEG Error Bounds for a Static Dipole Source with a Realistic Head Model Carlos H. Muravchik, Senior Member, IEEE, and Arye Nehorai,
More informationSVD-based optimal ltering with applications to noise reduction in speech signals Simon Doclo ESAT - SISTA, Katholieke Universiteit Leuven Kardinaal Me
Departement Elektrotechniek ESAT-SISTA/TR 999- SVD-based Optimal Filtering with Applications to Noise Reduction in Speech Signals Simon Doclo, Marc Moonen April, 999 Internal report This report is available
More informationFoundations of Computer Vision
Foundations of Computer Vision Wesley. E. Snyder North Carolina State University Hairong Qi University of Tennessee, Knoxville Last Edited February 8, 2017 1 3.2. A BRIEF REVIEW OF LINEAR ALGEBRA Apply
More informationA study of dipole localization accuracy for MEG and EEG using a human skull phantom
A study of dipole localization accuracy for MEG and EEG using a human skull phantom R. M. Leahy +, J. C. Mosher *, M. E. Spencer ++, M. X. Huang **, and J. D. Lewine *** + Signal & Image Processing Institute,
More informationOrder Selection for Vector Autoregressive Models
IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL. 51, NO. 2, FEBRUARY 2003 427 Order Selection for Vector Autoregressive Models Stijn de Waele and Piet M. T. Broersen Abstract Order-selection criteria for vector
More informationModelling temporal structure (in noise and signal)
Modelling temporal structure (in noise and signal) Mark Woolrich, Christian Beckmann*, Salima Makni & Steve Smith FMRIB, Oxford *Imperial/FMRIB temporal noise: modelling temporal autocorrelation temporal
More informationLinear Algebra Methods for Data Mining
Linear Algebra Methods for Data Mining Saara Hyvönen, Saara.Hyvonen@cs.helsinki.fi Spring 2007 The Singular Value Decomposition (SVD) continued Linear Algebra Methods for Data Mining, Spring 2007, University
More informationDirection of Arrival Estimation: Subspace Methods. Bhaskar D Rao University of California, San Diego
Direction of Arrival Estimation: Subspace Methods Bhaskar D Rao University of California, San Diego Email: brao@ucsdedu Reference Books and Papers 1 Optimum Array Processing, H L Van Trees 2 Stoica, P,
More informationLinear Algebra & Geometry why is linear algebra useful in computer vision?
Linear Algebra & Geometry why is linear algebra useful in computer vision? References: -Any book on linear algebra! -[HZ] chapters 2, 4 Some of the slides in this lecture are courtesy to Prof. Octavia
More informationDetection of Steady-State VEP s: Toward Electrophysiologic Measurement of Visual Acuity
Detection of Steady-State VEP s: Toward Electrophysiologic Measurement of Visual Acuity Carlos E. Davila Electrical Engineering Department Southern Methodist University Dallas, TX, USA 2 Collaborators
More informationNew Introduction to Multiple Time Series Analysis
Helmut Lütkepohl New Introduction to Multiple Time Series Analysis With 49 Figures and 36 Tables Springer Contents 1 Introduction 1 1.1 Objectives of Analyzing Multiple Time Series 1 1.2 Some Basics 2
More informationLinear and Non-Linear Responses to Dynamic Broad-Band Spectra in Primary Auditory Cortex
Linear and Non-Linear Responses to Dynamic Broad-Band Spectra in Primary Auditory Cortex D. J. Klein S. A. Shamma J. Z. Simon D. A. Depireux,2,2 2 Department of Electrical Engineering Supported in part
More informationOptimal Detection of Visual. Evoked Potentials. Medical Center, Dallas, Texas, USA. Fax: (214)
Optimal Detection of Visual Evoked Potentials Carlos E. Davila 1 Richard Srebro 2 Ibrahim A. Ghaleb 1 1 Electrical Engineering Department, Southern Methodist University, Dallas, Texas, USA 2 Depts. of
More informationPrincipal Component Analysis
Principal Component Analysis Introduction Consider a zero mean random vector R n with autocorrelation matri R = E( T ). R has eigenvectors q(1),,q(n) and associated eigenvalues λ(1) λ(n). Let Q = [ q(1)
More informationPrincipal Component Analysis
Principal Component Analysis CS5240 Theoretical Foundations in Multimedia Leow Wee Kheng Department of Computer Science School of Computing National University of Singapore Leow Wee Kheng (NUS) Principal
More informationIndependent Component Analysis
A Short Introduction to Independent Component Analysis with Some Recent Advances Aapo Hyvärinen Dept of Computer Science Dept of Mathematics and Statistics University of Helsinki Problem of blind source
More informationIndependent component analysis for noisy data MEG data analysis
Independent component analysis for noisy data MEG data analysis Shiro Ikeda PRESTO, JST Keisuke Toyama Shimadzu Inc. Abstract ICA (independent component analysis) is a new, simple and powerful idea for
More informationStatistical and Adaptive Signal Processing
r Statistical and Adaptive Signal Processing Spectral Estimation, Signal Modeling, Adaptive Filtering and Array Processing Dimitris G. Manolakis Massachusetts Institute of Technology Lincoln Laboratory
More informationAutomatic detection of the number of Raypaths
Automatic detection of the number of Raypaths Longyu Jiang, Jerome Mars To cite this version: Longyu Jiang, Jerome Mars. Automatic detection of the number of Raypaths. OCEANS MTS/IEEE Kona - Oceans of
More informationDistributed adaptive generalized eigenvector estimation of a sensor signal covariance matrix pair in a fully-connected sensor network
2 SIGNAL PROCESSING, VOL. 106, PP. 209-214, 2015 Distributed adaptive generalized eigenvector estimation of a sensor signal covariance matrix pair in a fully-connected sensor network Alexander Bertrand
More informationDOA Estimation of Uncorrelated and Coherent Signals in Multipath Environment Using ULA Antennas
DOA Estimation of Uncorrelated and Coherent Signals in Multipath Environment Using ULA Antennas U.Somalatha 1 T.V.S.Gowtham Prasad 2 T. Ravi Kumar Naidu PG Student, Dept. of ECE, SVEC, Tirupati, Andhra
More informationData Mining Lecture 4: Covariance, EVD, PCA & SVD
Data Mining Lecture 4: Covariance, EVD, PCA & SVD Jo Houghton ECS Southampton February 25, 2019 1 / 28 Variance and Covariance - Expectation A random variable takes on different values due to chance The
More informationNOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION. M. Schwab, P. Noll, and T. Sikora. Technical University Berlin, Germany Communication System Group
NOISE ROBUST RELATIVE TRANSFER FUNCTION ESTIMATION M. Schwab, P. Noll, and T. Sikora Technical University Berlin, Germany Communication System Group Einsteinufer 17, 1557 Berlin (Germany) {schwab noll
More informationLocalization of Multiple Deep Epileptic Sources in a Realistic Head Model via Independent Component Analysis
Localization of Multiple Deep Epileptic Sources in a Realistic Head Model via Independent Component Analysis David Weinstein, Leonid Zhukov, Geoffrey Potts Email: dmw@cs.utah.edu, zhukov@cs.utah.edu, gpotts@rice.edu
More informationPrincipal Components Analysis (PCA)
Principal Components Analysis (PCA) Principal Components Analysis (PCA) a technique for finding patterns in data of high dimension Outline:. Eigenvectors and eigenvalues. PCA: a) Getting the data b) Centering
More informationRecipes for the Linear Analysis of EEG and applications
Recipes for the Linear Analysis of EEG and applications Paul Sajda Department of Biomedical Engineering Columbia University Can we read the brain non-invasively and in real-time? decoder 1001110 if YES
More informationMAXIMUM A POSTERIORI ESTIMATION OF SIGNAL RANK. PO Box 1500, Edinburgh 5111, Australia. Arizona State University, Tempe AZ USA
MAXIMUM A POSTERIORI ESTIMATION OF SIGNAL RANK Songsri Sirianunpiboon Stephen D. Howard, Douglas Cochran 2 Defence Science Technology Organisation PO Box 500, Edinburgh 5, Australia 2 School of Mathematical
More informationDynamical MEG Source Modeling with Multi-Target Bayesian Filtering
r Human Brain Mapping 000:000 000 (009) r Dynamical MEG Source Modeling with Multi-Target Bayesian Filtering Alberto Sorrentino, 1 * Lauri Parkkonen,,3 Annalisa Pascarella, 1,4 Cristina Campi, 1,5 and
More informationPCA and admixture models
PCA and admixture models CM226: Machine Learning for Bioinformatics. Fall 2016 Sriram Sankararaman Acknowledgments: Fei Sha, Ameet Talwalkar, Alkes Price PCA and admixture models 1 / 57 Announcements HW1
More informationDETECTION of the number of sources measured by an
2746 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 58, NO 5, MAY 2010 Nonparametric Detection of Signals by Information Theoretic Criteria: Performance Analysis an Improved Estimator Boaz Nadler Abstract
More informationFinal Exam, Linear Algebra, Fall, 2003, W. Stephen Wilson
Final Exam, Linear Algebra, Fall, 2003, W. Stephen Wilson Name: TA Name and section: NO CALCULATORS, SHOW ALL WORK, NO OTHER PAPERS ON DESK. There is very little actual work to be done on this exam if
More informationA MULTIVARIATE MODEL FOR COMPARISON OF TWO DATASETS AND ITS APPLICATION TO FMRI ANALYSIS
A MULTIVARIATE MODEL FOR COMPARISON OF TWO DATASETS AND ITS APPLICATION TO FMRI ANALYSIS Yi-Ou Li and Tülay Adalı University of Maryland Baltimore County Baltimore, MD Vince D. Calhoun The MIND Institute
More informationNeural mass model parameter identification for MEG/EEG
Neural mass model parameter identification for MEG/EEG Jan Kybic a, Olivier Faugeras b, Maureen Clerc b, Théo Papadopoulo b a Center for Machine Perception, Faculty of Electrical Engineering, Czech Technical
More informationStatistics 910, #5 1. Regression Methods
Statistics 910, #5 1 Overview Regression Methods 1. Idea: effects of dependence 2. Examples of estimation (in R) 3. Review of regression 4. Comparisons and relative efficiencies Idea Decomposition Well-known
More informationRational Invariant Subspace Approximations with Applications
3032 IEEE TRANSACTIONS ON SIGNAL PROCESSING, VOL 48, NO 11, NOVEMBER 2000 Rational Invariant Subspace Approximations with Applications Mohammed A Hasan, Member, IEEE, Mahmood R Azimi-Sadjadi, Senior Member,
More informationPrincipal Component Analysis
Principal Component Analysis November 24, 2015 From data to operators Given is data set X consisting of N vectors x n R D. Without loss of generality, assume x n = 0 (subtract mean). Let P be D N matrix
More informationML ESTIMATION AND CRB FOR NARROWBAND AR SIGNALS ON A SENSOR ARRAY
2014 IEEE International Conference on Acoustic, Speech and Signal Processing ICASSP ML ESTIMATION AND CRB FOR NARROWBAND AR SIGNALS ON A SENSOR ARRAY Langford B White School of Electrical and Electronic
More information2-D SENSOR POSITION PERTURBATION ANALYSIS: EQUIVALENCE TO AWGN ON ARRAY OUTPUTS. Volkan Cevher, James H. McClellan
2-D SENSOR POSITION PERTURBATION ANALYSIS: EQUIVALENCE TO AWGN ON ARRAY OUTPUTS Volkan Cevher, James H McClellan Georgia Institute of Technology Atlanta, GA 30332-0250 cevher@ieeeorg, jimmcclellan@ecegatechedu
More informationAPP01 INDEPENDENT COMPONENT ANALYSIS FOR EEG SOURCE LOCALIZATION IN REALISTIC HEAD MODELS. Proceedings of 3ICIPE
Proceedings of 3ICIPE Inverse Problems in Engineering: Theory and Practice 3rd Int. Conference on Inverse Problems in Engineering June 3-8, 999, Port Ludlow, Washington, USA APP INDEPENDENT COMPONENT ANALYSIS
More informationLinear Algebra & Geometry why is linear algebra useful in computer vision?
Linear Algebra & Geometry why is linear algebra useful in computer vision? References: -Any book on linear algebra! -[HZ] chapters 2, 4 Some of the slides in this lecture are courtesy to Prof. Octavia
More informationSignal Modeling Techniques in Speech Recognition. Hassan A. Kingravi
Signal Modeling Techniques in Speech Recognition Hassan A. Kingravi Outline Introduction Spectral Shaping Spectral Analysis Parameter Transforms Statistical Modeling Discussion Conclusions 1: Introduction
More informationPrincipal Component Analysis CS498
Principal Component Analysis CS498 Today s lecture Adaptive Feature Extraction Principal Component Analysis How, why, when, which A dual goal Find a good representation The features part Reduce redundancy
More informationPCA & ICA. CE-717: Machine Learning Sharif University of Technology Spring Soleymani
PCA & ICA CE-717: Machine Learning Sharif University of Technology Spring 2015 Soleymani Dimensionality Reduction: Feature Selection vs. Feature Extraction Feature selection Select a subset of a given
More informationCollaborative Filtering: A Machine Learning Perspective
Collaborative Filtering: A Machine Learning Perspective Chapter 6: Dimensionality Reduction Benjamin Marlin Presenter: Chaitanya Desai Collaborative Filtering: A Machine Learning Perspective p.1/18 Topics
More informationUndercomplete Blind Subspace Deconvolution via Linear Prediction
Undercomplete Blind Subspace Deconvolution via Linear Prediction Zoltán Szabó, Barnabás Póczos, and András L rincz Department of Information Systems, Eötvös Loránd University, Pázmány P. sétány 1/C, Budapest
More informationBlock Bidiagonal Decomposition and Least Squares Problems
Block Bidiagonal Decomposition and Least Squares Problems Åke Björck Department of Mathematics Linköping University Perspectives in Numerical Analysis, Helsinki, May 27 29, 2008 Outline Bidiagonal Decomposition
More information