signal processing in neuroinformatics

9
Signal Processing in Neuroinformatics EEG Signal Processing Yongnan Ji

Upload: rose-anderson

Post on 31-Dec-2015

27 views

Category:

Documents


0 download

DESCRIPTION

Signal Processing in Neuroinformatics. EEG Signal Processing. Yongnan Ji. Modeling the EEG signal. Artifacts in the EEG. Nonparametric Spectral Analysis. Model-based spectral Analysis. Modeling the EEG signal. Stochastic. Deterministic VS. Nonlinear Modelling of EEG. - PowerPoint PPT Presentation

TRANSCRIPT

Page 1: Signal Processing in Neuroinformatics

Signal Processing in Neuroinformatics

EEG Signal Processing

Yongnan Ji

Page 2: Signal Processing in Neuroinformatics

Modeling the EEG signal.

Artifacts in the EEG.

Nonparametric Spectral Analysis.

Model-based spectral Analysis.

Page 3: Signal Processing in Neuroinformatics

Modeling the EEG signal.

Deterministic VS Stochastic

Linear Stochastic ModelsNonlinear Modelling

of EEG

ARMA, AR

Time-varing AR modelling

Multivariate AR modelling

AR modelling with impulse input

Page 4: Signal Processing in Neuroinformatics

Artifacts in the EEG.

Types of artifacts usually met

Eye movement and blinks, Muscle activity, Cardiac activity, Electrodes and equipment

Artifact Processing

Additive noise or multiplicative noise

How to deal with the artifact? Artifact Reduction Using Linear Filtering. Artifact Cancellation Using Linear Combined Reference

Signals. Adaptive Artifact Cancellation Using Linearly Combined

Reference Signals Artifact Cancellation Using Filtered Reference Signals

Page 5: Signal Processing in Neuroinformatics

Nonparametric Spectral Analysis

We can calculate an estimation of the power spectrum from the samples of the signal:

Mean and variance of the estimation changes against the selection of windows.

2. Spectral Parameters Spetral slope.

Hjorth descriptors.

Spectral Purity Index.

1. Fourier-based Power Spectrum Analysis

Page 6: Signal Processing in Neuroinformatics

Model-based Spectral Analysis

1. Variance of the input noise.

2. Methods to find the coefficients of the linear algorithm:

The Autocorrelation/Covariance Methods:

Minimization of the error variance. The Modified Covariance Method:

The variance is calcutated taking into acount backward prediction error.

Burg’s Method:

We explicitly make use of the recursion method. Estimation with lattice structure.

Page 7: Signal Processing in Neuroinformatics

Performance and Paramerters

3. Performance. Choosing method. Model order. Sampling rate.

4. Parameters.

Page 8: Signal Processing in Neuroinformatics

Exercise 3.7

)()()( xfxmxg

)()()( FMG

hz9

))2(sin()(

o

o

f

xfF F

Fourier Transform

)(G

9 09

99 0 810 108

Page 9: Signal Processing in Neuroinformatics

Exercise 3.7

)(M1

01

hz1),2sin()( mm fxfxM

Inverse-Fourier Transform