THE PROBLEM

04/19/23 ISSC DIT Kevin St. 1

THE PROBLEM

• TO CLASSIFY EEG SIGNALS USING WAVELET TRANSFORMS

AND NEURAL NETWORKS

04/19/23 ISSC DIT Kevin St. 2

• A SOLUTION

P ercen tag e R ecog n it ion

N eu ra l N e tworks

W ave le t C oeffic ien ts

W ave le t P acke t Tran s fo rm

P ercen tag e R ecog n it ion

N eu ra l N e tworks

W ave le t C oeffic ien ts

D isc re te W ave le t Tran s fo rm

M en ta l Task E E G F iles

04/19/23 ISSC DIT Kevin St. 3

The Brain

• A Neuron Cell

04/19/23 ISSC DIT Kevin St. 4

Electrode Placement

04/19/23 ISSC DIT Kevin St. 5

Time-Frequency Plane

y(t)=f(t)

N

n

NkjnenxkX1

)/2()()(

Discrete Fourier Transform

04/19/23 ISSC DIT Kevin St. 6

STFT f x t w t e d tfit( , ) ( ) ( )

2

Short Time Fourier Transform

f t. 1

4

Heisenberg Principle:

04/19/23 ISSC DIT Kevin St. 7

A ‘Sliding Window’

04/19/23 ISSC DIT Kevin St. 8

A Windowed EEG Signal

g t e et j to( ) ( ) ( ) 2

04/19/23 ISSC DIT Kevin St. 9

Spectrogram of the Mathematics task EEG signal

04/19/23 ISSC DIT Kevin St. 10

A 3-D Spectrogram of an EEG signal

04/19/23 ISSC DIT Kevin St. 11

Time-Frequency and Corresponding Basis Function

2/)( 2

)( a

t

a

tj

eeto

04/19/23 ISSC DIT Kevin St. 12

Logarithmic Tree

04/19/23 ISSC DIT Kevin St. 13

Wavelet Transform implementationusing Subband Coding

04/19/23 ISSC DIT Kevin St. 14

An example of a Pruning cost analysis

Prune if M(Parent)>M(child1)+M(child2)

04/19/23 ISSC DIT Kevin St. 15

Best Tree Structure for the Mathematics Task

04/19/23 ISSC DIT Kevin St. 16

The Wavelet Packet Transform

Time-Frequency Plane

04/19/23 ISSC DIT Kevin St. 17

Wavelet Coefficients: Augmented Time Format

04/19/23 ISSC DIT Kevin St. 18

Neural Network Structure

04/19/23 ISSC DIT Kevin St. 21

Recognition rate results for the WPT

04/19/23 ISSC DIT Kevin St. 22

Double-Sideband Suppressed Carrier:Reconstruction of the Wavelet Coefficients

04/19/23 ISSC DIT Kevin St. 23

A Designer Wavelet