3. Electroencephalography
Electroencephalography is a non-invasive brain recording
technique that records electrical activities in cortex
spontaneously .
Bandwidth: 0.05-128Hz; amplitude: 0.02- 0.4mV
(a) (b)
Figure 1: (a)Layers of human brain; and (b) Brain signal recording techniques. 3
4. 4
Delta: (< 4 Hz)
◦ adult slow-wave sleep
◦ in babies
◦ continuous-attention tasks
Theta: (4 - 8 Hz)
◦ higher in young children
◦ drowsiness in adults and teens
◦ idling
Alpha: (8 - 13 Hz)
◦ relaxed/reflecting
◦ closing the eyes
Beta: (14 - 30 Hz)
◦ active thinking, focus, high alert,
anxious
Gamma: (> 30 Hz)
◦ Displays during cross-modal sensory
processing
Mu: (7.5 - 12.5 Hz)
◦ Shows rest-state motor neurons
Gamma
5. History of EEG
• First human EEG was recorded by Hans Berger in 1924.
• Signal was recorded in semi non-invasive way .
• He is considered to be the father of EEG.
Figure 2: First recorded human electroencephalogram
5
12. Challenge: Motion Artifacts
Ambulatory EEG facilitates mobility making it prone to motion
Artifacts the most.
During motion, voltage fluctuations occur due to the
movements of electrode from their standard position or cable
sway.
Motion artifacts can occur from different sources like Ground
reaction forces, cyclic motion, head movements which are
related to daily activities.
It has the same frequency spectra ranging up to 50 Hz as EEG
signal and can have higher amplitude in millivolts.
The intensity and types of movements can vary person to
person making it most versatile artifacts. 12
13. Motivation
Motion artifacts intrude significant challenges on diagnosing
patients with critical neurological illness accurately.
The contaminated EEG signal can cause wrong detection,
prediction even miss the medical emergency.
Deterioration of BCI system performance during practical
applications
13
14. Literature review
14
Reference Dataset Activities Preprocessing
technique
Main
method
Performance
metrics
Novel Approaches for the
Removal of Motion Artifact
from EEG Recordings
Online By pulling
electrode
leads
N/A MTV(multiresoluti
on total
variation)and
MWTV(multiresol
ution weighted
total variation)
Average △SNR - 29.12dB(MTV) &
29.29db(MWTV)
Average η -
68.56% (MTV) & 67.51%(MWTV)
The Impact of Head
Movements on EEG and
Contact Impedance: An
Adaptive Filtering Solution for
Motion Artifact Reduction
Generated Head
movements
Stop band filter-
49-51Hz(3rd
order
butterworth
filter)
High pass filter-
0,194Hz(1st order
butterworth
filter)
Adaptive filtering MCAF result-high for D scores &
low for S scores (from the graph)
Motion artifact removal from
single channel
electroencephalogram signals
using singular spectrum
analysis
Online By pulling
electrode
leads
Bandpass
(0.5-30hz)
2nd order
butterworth
SSA-techniques Comparing with EEMD-CCA
computation complexity-six times
lesser
Percentage reduction-
11.388%
SNR-0.915dB
“The use of ensemble
empirical mode
decomposition with canonical
correlation analysis as a novel
artifact removal technique,”
Online By pulling
electrode
leads
NA ensemble
empirical mode
decomposition
with canonical
correlation
analysis
average ∆ SNR of 8.2 dB and
artifact was reduced by 52.3 %
yielding correlation of 0.63.
15. Literature review
Reference Dataset Activities Preprocessing
technique
Main method Performance metrics
EEG mobility artifact
removal for
ambulatory
eplieptic seizure
prediction
application
Generated Walking in
different speeds
bandpass(0.5-64hz)
Notch filter(50hz)
Infomax ICA Improvement in-
SNR(11dB)
RMSE(49mv)
Correlation-
77%
coherence-95%
For cleaned signal compared to
contaminated signal
Gaussian
Elimination-Based
Novel Canonical
Correlation Analysis
Method for EEG
Motion Artifact
Removal.
Online Seizure predicti
on and detection
from 23 pediatric
epilepsy patients
NA Cascaded method
EEMD-GECCA-SWT
Compared to CCA-
18% faster
accuracy-63.227%
Real-time Motion
Artifact Detection
and Removal for
Ambulatory BCI.
Generated Walking Kalman filtering
algorithm
ICA decomposition &
stride based template
regration
Accuracy - up to 93%
Removing Head
Movement Artifact
from EEG by ICA
Analysis and
Filtering
Generated Walking Bandpass(0.1-50hz)
4th order butterworth
filter
ICA with high pass
filter(3hz)
4th order butterworth
Higher theta and alpha power
than the ICA methods
Proposed Algorithm Generated Upper and lower
body
movements
PLI and offset
removal
Mother wavelet
transform “Haar” at
L=4
Motion artifact reduction 23.64%
15
16. Objective
To generate a reliable well defined dataset which confirms
the presence of motion artifact contaminated EEG signal
using EMOTIV along with a general algorithm to minimize
the effect of motion artifacts in EEG signals on MATLAB
platform.
16
33. Conclusion
Clean signal is not totally free from contamination.
To characterize the artifacts the duration of each artifact has
to be longer.
To characterize the artifact extracting the motion artifact from
EEG signal, controlled environment is a must.
Accelerometer does not record the same motion artifact
signal due to accelerometer placement.
Can not be depended on single device.
Different activities demands different approach to evaluate.
33
34. Future work
Artifact templates can be created artificially by extracting and
taking the envelope from neural signal for the ambulatory
activities. Then the artifact templates can be created into
different amplitude to do further study.
Using the datasets the same performance parameters can be
reevaluated for the existing paper related to Ambulatory EEG.
The data recorded without electrodes can be compared with
accelerometer data and study on this approach for new
papers.
Online extension tool can be created to check performance
parameter for any trial method by improving and including
number of trials for motion artifacts. 34