Artifact cancellation and nonparametric spectral analysis.

Slides:



Advertisements
Similar presentations
Biomedical Signal Processing
Advertisements

ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Periodograms Bartlett Windows Data Windowing Blackman-Tukey Resources:
Pattern Recognition and Machine Learning
Computer vision: models, learning and inference Chapter 13 Image preprocessing and feature extraction.
Let’s go back to this problem: We take N samples of a sinusoid (or a complex exponential) and we want to estimate its amplitude and frequency by the FFT.
Filtering Filtering is one of the most widely used complex signal processing operations The system implementing this operation is called a filter A filter.
Digital Signal Processing
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: The FIR Adaptive Filter The LMS Adaptive Filter Stability and Convergence.
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Newton’s Method Application to LMS Recursive Least Squares Exponentially-Weighted.
CSCE 641 Computer Graphics: Image Sampling and Reconstruction Jinxiang Chai.
Independent Component Analysis (ICA) and Factor Analysis (FA)
Image Enhancement.
Adaptive FIR Filter Algorithms D.K. Wise ECEN4002/5002 DSP Laboratory Spring 2003.
Spectral Analysis Spectral analysis is concerned with the determination of the energy or power spectrum of a continuous-time signal It is assumed that.
Introduction To Signal Processing & Data Analysis
AGC DSP AGC DSP Professor A G Constantinides 1 Digital Filter Specifications Only the magnitude approximation problem Four basic types of ideal filters.
Adaptive Signal Processing
Normalised Least Mean-Square Adaptive Filtering
Leakage & Hanning Windows
Introduction to Spectral Estimation
Introduction to Image Processing Grass Sky Tree ? ? Review.
Algorithm Taxonomy Thus far we have focused on:
Time-Domain Methods for Speech Processing 虞台文. Contents Introduction Time-Dependent Processing of Speech Short-Time Energy and Average Magnitude Short-Time.
Power Spectral Density Function
Filter Design Techniques
1 CS 551/651: Structure of Spoken Language Lecture 8: Mathematical Descriptions of the Speech Signal John-Paul Hosom Fall 2008.
T – Biomedical Signal Processing Chapters
Basics of Neural Networks Neural Network Topologies.
Speech Signal Representations I Seminar Speech Recognition 2002 F.R. Verhage.
Authors: Sriram Ganapathy, Samuel Thomas, and Hynek Hermansky Temporal envelope compensation for robust phoneme recognition using modulation spectrum.
Professor A G Constantinides 1 The Fourier Transform & the DFT Fourier transform Take N samples of from 0 to.(N-1)T Can be estimated from these? Estimate.
Chapter 6 Spectrum Estimation § 6.1 Time and Frequency Domain Analysis § 6.2 Fourier Transform in Discrete Form § 6.3 Spectrum Estimator § 6.4 Practical.
Real time DSP Professors: Eng. Julian S. Bruno Eng. Jerónimo F. Atencio Sr. Lucio Martinez Garbino.
§ 4.1 Instrumentation and Measurement Systems § 4.2 Dynamic Measurement and Calibration § 4.3 Data Preparation and Analysis § 4.4 Practical Considerations.
LEAST MEAN-SQUARE (LMS) ADAPTIVE FILTERING. Steepest Descent The update rule for SD is where or SD is a deterministic algorithm, in the sense that p and.
1 Spectrum Estimation Dr. Hassanpour Payam Masoumi Mariam Zabihi Advanced Digital Signal Processing Seminar Department of Electronic Engineering Noushirvani.
ECE 8443 – Pattern Recognition ECE 8423 – Adaptive Signal Processing Objectives: Derivation Computational Simplifications Stability Lattice Structures.
0 - 1 © 2007 Texas Instruments Inc, Content developed in partnership with Tel-Aviv University From MATLAB ® and Simulink ® to Real Time with TI DSPs Spectrum.
Chapter 11 Filter Design 11.1 Introduction 11.2 Lowpass Filters
Lecture#10 Spectrum Estimation
EEL 6586: AUTOMATIC SPEECH PROCESSING Speech Features Lecture Mark D. Skowronski Computational Neuro-Engineering Lab University of Florida February 27,
Professors: Eng. Diego Barral Eng. Mariano Llamedo Soria Julian Bruno
3.7 Adaptive filtering Joonas Vanninen Antonio Palomino Alarcos.
Autoregressive (AR) Spectral Estimation
Signal Processing in Neuroinformatics EEG Signal Processing Yongnan Ji.
Background 2 Outline 3 Scopus publications 4 Goal and a signal model 5Harmonic signal parameters estimation.
Outline Transmitters (Chapters 3 and 4, Source Coding and Modulation) (week 1 and 2) Receivers (Chapter 5) (week 3 and 4) Received Signal Synchronization.
Stochastic Process Theory and Spectral Estimation Bijan Pesaran Center for Neural Science New York University.
Vibrationdata 1 Power Spectral Density Function PSD Unit 11.
Principal components analysis (PCA) as a tool for identifying EEG frequency bands: I. Methodological considerations and preliminary findings Jürgen Kayser,
A. R. Jayan, P. C. Pandey, EE Dept., IIT Bombay 1 Abstract Perception of speech under adverse listening conditions may be improved by processing it to.
Stochastic Process Theory and Spectral Estimation
Power Spectral Estimation
SYSTEMS Identification Ali Karimpour Assistant Professor Ferdowsi University of Mashhad Reference: “System Identification Theory.
Performance of Digital Communications System
Professor A G Constantinides 1 Digital Filter Specifications We discuss in this course only the magnitude approximation problem There are four basic types.
1 C.A.L. Bailer-Jones. Machine Learning. Data exploration and dimensionality reduction Machine learning, pattern recognition and statistical data modelling.
Medical Image Analysis
Speech Enhancement Summer 2009
Computational Data Analysis
A New Technique for Sidelobe Suppression in OFDM Systems
CS 591 S1 – Computational Audio
Spectral Analysis Spectral analysis is concerned with the determination of the energy or power spectrum of a continuous-time signal It is assumed that.
The Chinese University of Hong Kong
Image Analysis Image Restoration.
A Tutorial on Bayesian Speech Feature Enhancement
Chapter 7 Finite Impulse Response(FIR) Filter Design
Volume 34, Issue 5, Pages (May 2002)
Chapter 7 Finite Impulse Response(FIR) Filter Design
Feature Selection in BCIs (section 5 and 6 of Review paper)
Presentation transcript:

Artifact cancellation and nonparametric spectral analysis

Outline  Artifact processing  Artifact cancellation  Nonparametric spectral analysis

Introduction  Artifact processing Rejection  cancellation Rejection  cancellation Rejection main alternative Rejection main alternative one would hope to retain dataone would hope to retain data Cancellation requirements Cancellation requirements clinical informationclinical information no new artifactsno new artifacts spike detectorsspike detectors Additive/multiplicative model Additive/multiplicative model Artifact reduction using linear filtering Artifact reduction using linear filtering

Artifact cancellation  Using linearly combined reference signals  Adaptive artfact cancellation using linearly combined reference signals  Using filtered reference signals

Linearly combined reference signals  Eye movements & blinks several referene signals several referene signals positioning positioning additive model additive model EOG linearly trasferred to EEG EOG linearly trasferred to EEG weightsweights

In detail  Uncorrelated  Mean square error  Minimization, differentation  Spatial correlation, cross correlation fixed over time fixed over time zero gradient zero gradient  Estimation blinks, eye-movements at onset blinks, eye-movements at onset

In detail 2  Number of reference signals  Only EOG cancelled  ECG  Rejection used a lot (in MEG) expect when lots of blinks (ssp) expect when lots of blinks (ssp)

Adaptive version  Time-varying changes  Tracking of slow changes  Adaptive algorithm LMS LMS weight(s) function of time weight(s) function of time optimal solution changes with timeoptimal solution changes with time method of steepest descent method of steepest descent negative error gradient vector negative error gradient vector

In detail  Parameter selection time time noise noise  Expectation instantaneous value instantaneous value zero setting zero setting performance estimation performance estimation fluctuation of weights fluctuation of weights

Filtered reference signals  EOG potentials exhibit frequency dependence in trasfer to EEG sensor through tissue in trasfer to EEG sensor through tissue blinks and eye movements blinks and eye movements  Improved cancellation with transfer function replacement spatial and temporal information spatial and temporal information v 0 estimation v 0 estimation FIR (lengths) FIR (lengths)

Details  Stationary processes Second order characterisrics Second order characterisrics Correlation information fixed Correlation information fixed

Details 2  No a priori information can be implemented, modified error can be implemented, modified error  Also adaptive version exists a priori impulse responses calculated at calibration a priori impulse responses calculated at calibration

Nonparametric spectral analysis  Richer characterization of background activity that with 1D histograms  EEG rhythms  Correlate signals with sines and cosines  When? Gaussian stationary signals Gaussian stationary signals Stationary estimatationStationary estimatation Normal spontaneous waking activity Normal spontaneous waking activity

Nonparametric 2  Fourier-based power spectrum analysis no modeling assumptions no modeling assumptions  Spectral parameters interpretation interpretation

Fourier-based power spectrum analysis  Power spectrum characterized by correlation function (stationary) If ergodic, approximate with time average estimator (negative lags) If ergodic, approximate with time average estimator (negative lags) combination called periodogram combination called periodogram equals squared magnitude of DFT equals squared magnitude of DFT

Fourier considerations  Periodogram biased window dependent (convolution) window dependent (convolution) smearing (main lobe) smearing (main lobe) leakage (side lobes) leakage (side lobes) synchronized rhythm better described by power in frequency bandsynchronized rhythm better described by power in frequency band variance periogoram variance periogoram does not approach zero with sample increasedoes not approach zero with sample increase consistency consistency

Periodogram  Windowing and averaging leakage & periodogram variance reduction leakage & periodogram variance reduction  Windows from rectangular to smaller sidelobes from rectangular to smaller sidelobes wider main lobe, spectral resolutionwider main lobe, spectral resolution  Variance reduction nonoverlapping segments, averaging nonoverlapping segments, averaging resolution decrease, trade-offresolution decrease, trade-off combinations, degree of overlapcombinations, degree of overlap

And then what...

Spectral parametrs  Resulting power spectrum often not readilty interpreted Condensed into compact set of parameters Condensed into compact set of parameters feature extraction feature extraction parameters describing prominent features of the spectrumparameters describing prominent features of the spectrum peaks, frequencies peaks, frequencies general usagegeneral usage

Spectral choices  Visual inspection format selection format selection assessing represantiveness assessing represantiveness  Scaling scope of the analysis scope of the analysis

Parameters  Power in frequency bands  Peak frequency  Spectral slope  Hjort descriptors  Spectral purity index

Power in frequency bands  Fixed/statistical bands alpha, beta, theta etc. alpha, beta, theta etc. from data from data  Ratio of, absolute power comparison, nonphysiological factors comparison, nonphysiological factors

Peak frequency  Frequency, amplitude, width  ad hoc methods for determining peaks  more than just maximum median, mean median, mean

Spectral slope  EEG activity made of 2 component rhythmic, unstructured rhythmic, unstructured  Based on decay of high frequency components one parameters approximation one parameters approximation least squares errorleast squares error  Quantifcation of EEG  Preconditioning of power estimate

Hjort descriptors  Spectral moments H 0 (activity) H 0 (activity) H 1 (mobility) H 1 (mobility) H 2 (complexity) H 2 (complexity)  Signal power, dominant frequency, bandwidth  Effectively in time domain  Clinically useful

Spectral purity index (SPI)  Heuristic  Reflects signal bandwidth (H 2 )  How well signal is described by a single frequency noise susceptibility noise susceptibility

Summary  Artifact cancellation reference signals reference signals linear combinations, filtering linear combinations, filtering adaptive version(s)adaptive version(s)  Spectral parameters nonparametric nonparametric no modellingno modelling parametric parametric interpretationinterpretation