Upload
michele-barsotti
View
323
Download
1
Embed Size (px)
DESCRIPTION
Workshop: "Brain Computer Interfaces & Haptics" Haptics Symposium 2014, Houston Texas
Citation preview
Haptics
Symposium
2014
Haptics
Symposium
2014
Exoskeleton control by Motor Imagery
BCI for upper limb neurorehabilitation
Michele Barsotti, Antonio Frisoli
[email protected]@sssup.it
Brain Computer Interfaces
&
Haptics
February 23-26, 2014,
Houston TexasHaptics Symposium
2014
Haptics
Symposium
2014
Haptics
Symposium
2014
• Brain activity acquisition methods
• General Brain Computer Interfaces
• Brain anatomy
• Movements EEG correlates
• Decoding movement intention byanalyzing EEG
• Feedbacks for motor-imagery-BCI
• Implementing a MI-BCI paradigm
OUTLINE
Haptics
Symposium
2014
Haptics
Symposium
2014
BRAIN ACTIVITY ACQUISITION
Haptics
Symposium
2014
Haptics
Symposium
2014
ACQUIRING BRAIN ACTIVITY
Temporal Resolution [s]
Sp
ati
al
Reso
luti
on
[cm
]
BASED ON THE
BLOOD FLOW
VARIATION
BASED ON THE
MAGNETIC-
ELECTRICAL
ACTIVITY
Haptics
Symposium
2014
Haptics
Symposium
2014
ACQUIRING BRAIN ACTIVITY
Temporal Resolution [s]
Sp
ati
al
Reso
luti
on
[cm
]
ElectroCorticoGraphy
(ECoG)
Very good spatial and
temporal resolution (firing
of a single neuron)
INVASIVE
Surgical intervention
Haptics
Symposium
2014
Haptics
Symposium
2014
ACQUIRING BRAIN ACTIVITY
Temporal Resolution [s]
Sp
ati
al
Reso
luti
on
[cm
]
• Most widely used strategy for BCI applications
• Good Temporal
Resolution
• Several portable, cheap systems exist
•Motion artifacts and interferences can be greatly reduced by employing active electrodes
EEG is the record of
electrical activity of brain by
placing the electrodes on
the scalp.
Haptics
Symposium
2014
Haptics
Symposium
2014
EEG SIGNALs FEATURES
AMPLITUDE RANGE:
Wake EEG:: Vpp = 100µV
Sleep EEG: Vpp = 300µV
FREQUENCY RANGE:
From 0.01 to 100 Hz
COMMON EEG ARTIFACTs:
Eye blinking (eye movement)
Muscular activity (EMG)
Ambient Noise + (50Hz-60Hz)
Electrodes Movement
Zero Mean
Haptics
Symposium
2014
Haptics
Symposium
2014
NATURAL EEG RHITMIC ACTIVITYBand [Hz] Normaly Location
Gamma 32 + Displays during cross-modal sensory
processing and short-term memory
Somatosensory cortex
Beta 16 -
32
active thinking, focus, hi alert,
anxious
both sides, symmetrical
distribution, most evident
frontally; low-amplitude waves
Alpha 8 -
16
relaxed/reflecting
closing the eyes
inhibition control
posterior regions of head, both
sides, higher in amplitude on
non-dominant side.
Mu 8 -
12
Shows rest-state motor neurons Sensorimotor cortex
Theta 4 - 8 higher in young children
drowsiness in adults and teens
idling
Found in locations not related to
task at hand
Delta up to
4
adult slow-wave sleep
Has been found during some
continuous-attention tasks
frontally in adults, posteriorly in
children; high-amplitude waves
Haptics
Symposium
2014
Haptics
Symposium
2014
NATURAL EEG RHITMIC ACTIVITYBand [Hz] Normaly Location
Gamma 32 + Displays during cross-modal sensory
processing and short-term memory
Somatosensory cortex
Beta 16 -
32
active thinking, focus, hi alert,
anxious
both sides, symmetrical
distribution, most evident
frontally; low-amplitude waves
Alpha 8 -
16
relaxed/reflecting
closing the eyes
inhibition control
posterior regions of head, both
sides, higher in amplitude on
non-dominant side.
Mu 8 -
12
Shows rest-state motor neurons Sensorimotor cortex
Theta 4 - 8 higher in young children
drowsiness in adults and teens
idling
Found in locations not related to
task at hand
Delta up to
4
adult slow-wave sleep
Has been found during some
continuous-attention tasks
frontally in adults, posteriorly in
children; high-amplitude waves
Haptics
Symposium
2014
Haptics
Symposium
2014
GENERAL BRAIN COMPUTER
INTERFACES
Haptics
Symposium
2014
Haptics
Symposium
2014
GENERAL BCI FRAMEWORK
SIGNAL PROCESSING
FEATURES EXTRACTION
1 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
-0.2 -0.1 0 -0.2 -0.1 0
CSP
13 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4VAR
MAXVAR
MIN
SIGNAL CLASSIFICATION
APPLICATION OUTPUT
BIOFEEDBACK
USER MENTAL
STRATEGY
BRAIN SIGNALs
ACQUISITION
Haptics
Symposium
2014
Haptics
Symposium
2014
Feedback for
subject training
Machine
learning
BCIs represent a set of techniques to allow direct control of a software or device via brain activity – without the need of a motor output
The most common BCI approach exploits voluntary modulation of EEG activity, although more invasive approaches have been explored
These techniques have successfully been employed to aid disabled patients
Recently BCIs have also been investigated as a rehabilitation tool
GENERAL BCI FRAMEWORK
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI CATEGORIES
INVASIVE NON-INVASIVE
Without penetrating the
skalp, mostly EEG, rarely
magnetoencephalogram
(MEG) or functional
magnetic resonance
imaging fMRI
- Several portable, cheap systems
exist
- Motion artifacts and interferences
can be greatly reduced by employing
active electrodes
DEPENDING ON THE ACQUISITION SYSTEM
Implanted sensors
(electrode array, needle
electrodes,
electrocorticogram ECoG)
-Control of 2-3 DoF, with good
accuracy.
-Implants have only been tested for
months after surgery
--Highly expensive
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI
Invasive Non invasive
Single recording site
Multiple recording sites
ECoG
EEG
MEG
fMRI
Classification: signal acquisition
Haptics
Symposium
2014
Haptics
Symposium
2014
Insertion of arrays of microelectrodes in cortical tissue
Control of 2-3 DoF, with good accuracy.
Implants have only been tested for months
after surgery
Highly expensiveHochberg et al., Nature, 2006
Invasive vs. non-invasive BCI
Invasive BCI
Non-invasive BCI
EEG systems range from low to high density (2 to 256 eletrodes)
Several portable, cheap systems exist
Motion artifacts and interferences can be
greatly reduced by employing active electrodes
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI CATEGORIES
INDEPENDENT DEPENDENT
A Dependent BCI does not
use the brain’s normal
output pathways to carry
the message, but activity in
these pathway is needed to
generate the brain activity
that does carry it.
Independent from
peripheral nerves and
muscles, using only
central nervous system
(CNS) activity
DEPENDING ON THE MENTAL STRATEGY
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI CATEGORIES
ENDOGENOUS EXOGENOUS
Evoked Potentials:
Users modulate brain
responses to external
stimuli
SSVEP
p300
Unstimulated Brain
Signals:
Users can voluntarily
produce the required
signals
(Motor Imagery,
Computational Task)
DEPENDING ON THE MENTAL STRATEGY
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI CATEGORIES
ASYNCHRONOUS
Commands can only be
emitted synchronously with
external pace.
The system detects
when the user wants to
emit a command
DEPENDING ON THE COMMAND-TIMING
SYNCHRONOUS
The differences in EEG response following different stimuli are
used to discriminate what subjects want
Subjects are asked to perform visual imagery tasks and
the local changes in EEG power spectra are recorded
Haptics
Symposium
2014
Haptics
Symposium
2014
SSVEP
VEP
MOTOR IMAGERY ERP (i.e.P300)
BCI CATEGORIES - SUMMARY
EXOGENOUSENDOGENOUSD
EPEN
DEN
TIN
DEPEN
DEN
T
Haptics
Symposium
2014
Haptics
Symposium
2014
BRAIN ANATOMY & EEG MOVEMENTS
CORRELATES
Haptics
Symposium
2014
Haptics
Symposium
2014
BRAIN ANATOMY
[Martini, 2006]
Haptics
Symposium
2014
Haptics
Symposium
2014
The Primary Somatic Sensory Cortex (Parietal Lobe) and the Primary Motor
Cortex (Temporal Lobe) are the most important regions for BCI research.
I
III
IV V
IIBRAIN ANATOMY: THE CEREBRAL CORTEX
Haptics
Symposium
2014
Haptics
Symposium
2014
The Primary Somatic Sensory Cortex (Parietal Lobe) and the Primary Motor
Cortex (Temporal Lobe) are the most important regions for BCI research.
I
III
IV V
IIBRAIN ANATOMY: THE CEREBRAL CORTEX
M1 S1
Haptics
Symposium
2014
Haptics
Symposium
2014
The Primary Somatic Sensory Cortex (Parietal Lobe) and the Primary Motor
Cortex (Temporal Lobe) are the most important regions for BCI research.
I
III
IV V
IIBRAIN ANATOMY: THE CEREBRAL CORTEX
M1 S1
Haptics
Symposium
2014
Haptics
Symposium
2014
TYPES OF MOVEMENTThree types of movements may occur in respect of to ascending
and descending signals via different pathways and at different
levels: • Reflexes movement : are performed
subconsciously and can occur at an exclusively
spinal level
• Rhythmic movement: stereotyped action
involving repetitions of the same movements
The control is at the spinal level without
involvement of higher cortical control
• Voluntary movement: usually goal directed
and therefore fully conscious. It arises in the
motor cortex and is executed by the spinal
cord.
Haptics
Symposium
2014
Haptics
Symposium
2014
TYPES OF MOVEMENTThree types of movements may occur in respect of to ascending
and descending signals via different pathways and at different
levels: • Reflexes movement : are performed
subconsciously and can occur at an exclusively
spinal level
• Rhythmic movement: stereotyped action
involving repetitions of the same movements
The control is at the spinal level without
involvement of higher cortical control
• Voluntary movement: usually goal directed
and therefore fully conscious. It arises in the
motor cortex and is executed by the spinal
cord.
When a voluntary movement is started, neurons in the
M1 send commands to upper and lower motor neurons.
The M1 needs to be stimulated by neurons from the
premotor cortex and the supplementary motor area
(SMA), which support and coordinate the M1, in order to
initiate a voluntary movement
Haptics
Symposium
2014
Haptics
Symposium
2014
Motor imagery is a mental process by which
an individual rehearses or simulates a given
action.
MOTOR IMAGERY
Performing motor imagery or attempting a
movement (i.e. for patients) influences the
brain activity as the voluntary movements do.
Haptics
Symposium
2014
Haptics
Symposium
2014
Why MOTOR IMAGERY is suitable for BCI?
• No need of external stimulus (it could be asynchronous)
• Not depend in any way on the brain’s normal output/input pathways
(independent)
• Possibility to provide different commands depending on which body
part is evolved in the simulated action
• Mental practice of motor actions via BCI training affect neuro-
rehabilitation in a positive way.
• the power in μ (8-12 Hz) and β (12-24Hz) EEG rhythms are affected
by motor imagery: Event Related Spectral Perturbation (ERSP)
• Users learn to perform motor imagery tasks
• Can be employed event if the motor areas are impaired
• Works mostly for digital control, has a fast response
Haptics
Symposium
2014
Haptics
Symposium
2014
DECODING MOVEMENT
INTENTIONS BY ANALIZING EEG
Haptics
Symposium
2014
Haptics
Symposium
2014
Event Related Potential (ERP):
- Repeatedly present discrete stimulus, average
raw EEG responses across presentations.
Characteristic feature (eg. P300)
Event Related Spectral Perturbation (ERSP):
Frequency band changes
- Average spectral features across presentation.
- Characteristic suppression/increase in power
(ERD/ERS: Event Related De-Synchronization).
EEG PHENOMENAL USABLE FOR BCI
Event Related Spectral Perturbation (ERSP) and Event
Related Potential ERP are the measured brain response that
are the direct result of a specific sensory, cognitive, or motor
event.
Haptics
Symposium
2014
Haptics
Symposium
2014
time
Fre
que
ncy [
Hz]
epochtimechannel X epochfrequencytimechannel X
AVERAGING
time
Am
plit
ude [µ
V]
ERP ERSP
• The ERPs and ERSP should be extracted from the background
noise mediating many recordings (Epochs or Trials)
Haptics
Symposium
2014
Haptics
Symposium
2014
Am
plit
ude
[µV
]
EEG background
noise ~ 1/sqrt(N)
Costant Signal ERP
Repetition (N)
Post-Stimulus
EEG
Costant
Signal
Background
Noise
average
ERP
average
Signal
average
Noise
AVERAGING THEORYS/N ratio increases as a function of the square root of the number of trials.
Haptics
Symposium
2014
Haptics
Symposium
2014
MOTOR IMAGERY CORRELATES IN EEGPerforming (or imagining) a motor action influences the EEG with two
main phenomena:
Haptics
Symposium
2014
Haptics
Symposium
2014
SLOW CORTICAL POTENTIALS [Kornhuber and Deecke (1965) ]
• Know as BereitschaftPotential
(readiness potential) or Movement
Related Cortical Potentials (MRCPs).• Slow oscillations preceding the
movement
• Localized over the supplementary
motor area (SMA)
• Steps for MRCP detection
• Spatial filter,
• LP frequency filter
• Template extraction from the
training data
• matching with the ongoing eeg
MOTOR IMAGERY CORRELATES IN EEG
•Frequency close to the DC -> very
challenging to detect in single trial
Haptics
Symposium
2014
Haptics
Symposium
2014
SENSORIMOTOR RHYTHMS [Pfurtscheller and Lopes da Silva, (1999)]
• the power in μ (8-12 Hz) and β (12-24
Hz) EEG rhythms are affected by motor
imagery.
•Know also as Event Related
De/Synchronization (ERD,ERS)
•Steps for MRCP detection
• Spatial filter,
• Band Pass frequency filter
• Feature extraction
• LDA classifier
• High average classification
accuracy (>80%)
MOTOR IMAGERY CORRELATES IN EEG
Haptics
Symposium
2014
Haptics
Symposium
2014
ERD extraction: example with motor imageryCollecting Trials
from a specific
electrode
Bandpass on the
specific frequency
Squaring Signals
Averaging over
Trials
Smoothing
[Pfurtscheller and Lopes da Silva, (1999)]
Haptics
Symposium
2014
Haptics
Symposium
2014
β ERD
13-30 Hz
µ ERD
8-12 Hz
Event Related
De\Synchronization
ERD Motor Imagery of
right hand
movement
EVENT RELATED SPECTRAL PERTURBATION
SENSORIMOTOR RHYTHMS
Haptics
Symposium
2014
Haptics
Symposium
2014
MOTOR IMAGERY: SIGLE TRIAL DETECTIONThe important features of the motor imagery are:
The frequency band.
The spatial localization
A priori knowledgment:
The frequency band are mu (8 -13Hz) and beta (15-30 Hz).
The spatial localization is over the sensory motor
Very high intersubject
variability!
Need of optimized
spatial filters
Haptics
Symposium
2014
Haptics
Symposium
2014
The aim of spatial filtering is to improve the signal-to-noise ratio by
creating a virtual channel which is a (linear, in the following cases)
combination of the input channels of the filter.
A spatial filters can optimize the data extracted from an high number of
electrodes reducing the dimension of the features'space to only few
significant dimensions.
N-channel input (ex. 16 ch) 1-optimized channel output
1 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
-0.2 -0.1 0 -0.2 -0.1 0
CSP
13 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
SPATIAL FILTERING
y(t) = a*ch1(t) + b*ch2(t) ....
Haptics
Symposium
2014
Haptics
Symposium
2014
Optimized Spatial filter: Common Spatial Pattern – CSP
VAR MAX VAR MIN
VAR MIN
RAW
CHANNELS
FIRST AND LAST
CSP FILTER
PROJECTED
DATA
REST MOVERESTMOVE
Trial i Trial i+1 Trial i Trial i+1
VAR MAX
[Pfurtscheller 1999]
Common Spatial Pattern (CSP) is a supervised spatial filtering
method for two-class discrimination problems, which finds
directions that maximize variance for one class and at the same
time minimize variance for the other class.
1 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
-0.2 -0.1 0 -0.2 -0.1 0
CSP
13 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
1 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
-0.2 -0.1 0 -0.2 -0.1 0
CSP
13 CSP
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
Haptics
Symposium
2014
Haptics
Symposium
2014
WHITENING
MATRIX
TRIALS
CLASS A
TRIALS
CLASS B
COVARIANCE
CLASS A
COVARIANCE
CLASS B
i
Ti
A
i
A
Ti
A
i
AA
XXtrace
XXR
)(
i
Ti
B
i
B
Ti
B
i
BB
XXtrace
XXR
)(
BAc RRR
T
CCCC UUR
T
CC UW1
T
AA WWRS T
BB WWRS T
AA UUS T
BB UUS
IBA
WUP T
PXZ
COMPOSITE
COVARIANCE
Transformed
Covariance A
Transformed
Covariance BEIGENVECTOR
PROJECTION
MATRIX
EIGENVALUES
Common Spatial Pattern – Algorithms
Haptics
Symposium
2014
Haptics
Symposium
2014
• The scalp-plot of the Common Spatial Pattern can be also used to give a physiological interpretation of the data
Common Spatial Pattern: advantages
• Since variance of band-pass filtered signals is equal to band-power, CSP filters are well suited to discriminate mental states characterized by spectral perturbations (ERD and motor imagery based BCIs).
Haptics
Symposium
2014
Haptics
Symposium
2014
The log-scaled band-power values in the mu and beta band of
the resulting two projected channels, can be used as a two-
dimensional feature of the brain activity. Classification is
performed using a linear discriminant classifier (LDA) or a
support vector machine (SVM)
CLASSIFICATION
Haptics
Symposium
2014
Haptics
Symposium
2014
CSP VARIANTS
CSP – Pfurtscheller 1998
FWM – Liu 2010
CSSSP – Blankertz 2006
CSSP – Lemm 2005
SPEC-CSP – Tomioka 2006
SB-CSP – Novi 2008
FB-CSP – Ang 2008
dCSP – Wang 2010
SSCSP – Arvaneh 2011
I-CSP – Blankertz 2008
With
Frequency
Optimization
Furhter Spatial
Optimization
Haptics
Symposium
2014
Haptics
Symposium
2014
FEEDBACKs for
motor imagery -
BCI
Haptics
Symposium
2014
Haptics
Symposium
2014
FEEDBACK FOR MOTOR IMAGERYThe biofeedback provided as a response to the mental activity can
improves the usability of motor imagery BCI.
The congruency of the provided feedback with the mental task is
expected to ease the performance of motor imagery.
Game
Illusion
Virtual reality
Exoskeleton
VISUAL
PROPRIOCEPTIVE
Haptics
Symposium
2014
Haptics
Symposium
2014
motor imagery –
BCI in
neurological
rehabilitation
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI in neurological rehabilitation
Daly & Wolpaw, Lancet, 2008
Haptics
Symposium
2014
Haptics
Symposium
2014
Daly & Wolpaw, Lancet, 2008Goal: The subject should be able to control
muscle activity through brain activity
BCI in neurological rehabilitation
Haptics
Symposium
2014
Haptics
Symposium
2014
Daly & Wolpaw, Lancet, 2008
Strategy 1: Train subjects to modulate brain activity via visualization and voluntary control of relevant features
BCI in neurological rehabilitation
Haptics
Symposium
2014
Haptics
Symposium
2014
Daly & Wolpaw, Lancet, 2008
Strategy 2: Train subjects by using brain activity to aid motion with assistive devices
BCI in neurological rehabilitation
Haptics
Symposium
2014
Haptics
Symposium
2014
A new multimodal architecture for gaze-independent brain–computer
interface (BCI)-driven control of a robotic upper limb exoskeleton for
stroke rehabilitation to provide active assistance in the execution of
reaching tasks in a real setting scenario.
Object
1
Object
2
Work
plane
Kinect
Eye tracker
BC
I
Haptics
Symposium
2014
Haptics
Symposium
2014
VAR
MIN
OPTIMAL CHANNELSMOVE
VAR
MAX
REST MOVE REST
VAR
MAX
VAR
MIN
ORIGINAL CHANNELSCSP
FILTERS
SVM
CLASSIFIER
TRAINING PHASE
VISUAL CONDITION ROBOT CONDITION
Involving the BCI module only and
the visual feedback of a virtual arm
controlled through motor
The subject performed a test session
with the complete system:
Kinect – EyeTracker – BCI – ArmExos
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI-REHABILITATION PROTOCOL
BCIEEG acquisition
& processing
L-EXOSproprioceptive
feedback
MONITORvisual feedback
• TRAINING PHASE: visual and proprioceptive feedback are provided accordingly to the task
• EXERCISE PHASE: the real-time classification output of the BCI was used for driving the
proprioceptive and visual feedback
ALL PATIENTS WERE ABLE TO CONTROL THE
BCI SYSTEM AFTER THE FIRST TWO
SESSION
5 right hemiparetic stroke
patients enrolled
SESSION STRUCTURE:
• MOVEMENT: the patient have to perform motor imagery of his impaired arm
• REST: the patient have to hold a resting mental state
TASKs REQUIRED:
Haptics
Symposium
2014
Haptics
Symposium
2014
BCI paradigm
based on Motor
Imagery
Haptics
Symposium
2014
Haptics
Symposium
2014
EEG
acquisition
Signal filtering
and
conditioning
Features
extraction
Features
classification
Online operations:
User
Offline BCI training
Frequency
bands and
artifact
removal
parameters
Spatial
Filter
parameters
Classifier
weights
Real-time feedback
MOTOR IMAGERY BCI: WORKFLOW
Haptics
Symposium
2014
Haptics
Symposium
2014
EEG CONFIGURATION
EEG channels: minimal configuration
Frontal ground electrode
Reference ear lobe electrode
Electrodes covering the motor cortex
Electrode for eye-blink detection and removal
Feature extraction
The power in the mu (8-12 Hz) and beta (16-24 Hz) bands is computed over 500 ms windows.
Haptics
Symposium
2014
Haptics
Symposium
2014
TRAINING PHASE
Training paradigmSubjects are asked to perform several motor imagery trials.
1. Feature classificationAcquired data is classified into two or more classes via machine
learning techniques, to optimize feature classification
2. Subject trainingThe subject is trained again with the output of the feature classifier as a
feedback signal, in order to optimize its motion imagery
TRIAL STRUCTURE
Haptics
Symposium
2014
Haptics
Symposium
2014
DATA PROCESSINGTRAINING
• Import data with the channel location
• Subdivide data into epochs for the two classes
• Remove artifactuated epochs
• Train the Common Spatial filter
• Extract Features
• Train the classifier
it is possible to predict the
BCI performance by a
visual inspection of both
the time-frequency plot of
the CSP-projected
channels and the
features plot
Haptics
Symposium
2014
Haptics
Symposium
2014
DATA PROCESSING: Visual Inspection
Haptics
Symposium
2014
Haptics
Symposium
2014
Time Frequency plot raw channels
Time [ms]
Fre
qu
en
cy [H
z]
C3
-2000 0 2000 4000
10
20
30
-2
0
2
Time [ms]
Fre
qu
en
cy [H
z]
CZ
-2000 0 2000 4000
10
20
30
-2
0
2
-2
0
2
CHANNELS ERD MAPS - MOVE
Time [ms]
Fre
qu
en
cy [H
z]
C4
-2000 0 2000 4000
10
20
30
Time [ms]
Fre
qu
en
cy [H
z]
C3
-2000 0 2000 4000
10
20
30
-2
-1
0
1
2
Time [ms]
Fre
qu
en
cy [H
z]
CZ
-2000 0 2000 4000
10
20
30
-2
-1
0
1
2
-2
-1
0
1
2
CHANNELS ERD MAPS - REST
Time [ms]
Fre
qu
en
cy [H
z]
C4
-2000 0 2000 4000
10
20
30
Click on electrodes to toggle name/number
Click on electrodes to toggle name/number
Click on electrodes to toggle name/number
Haptics
Symposium
2014
Haptics
Symposium
2014
Time Frequency plot CSP projected channels
Time [ms]
Fre
qu
en
cy [H
z]
MOVE - First CSP
-2000 -1000 0 1000 2000 3000 4000
10
20
30
-5
0
5
Time [ms]
Fre
qu
en
cy [H
z]
MOVE - Last CSP
-2000 -1000 0 1000 2000 3000 4000
10
20
30
-2
0
2
Time [ms]
Fre
qu
en
cy [H
z]
REST - First CSP
-2000 -1000 0 1000 2000 3000 4000
10
20
30
-2
0
2
-2
0
2
Time [ms]
Fre
qu
en
cy [H
z]
REST - Last CSP
-2000 -1000 0 1000 2000 3000 4000
10
20
30
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
FC3 FCZ FC4
C5 C3 C1 CZ C2 C4 C6
CP3 CPZ CP4
First CSP
First CSP
Last CSP
Last CSP
REST trials
MOVE trials
Haptics
Symposium
2014
Haptics
Symposium
2014
0 1000 2000 30000
20
40
60
80
100
Time [ms]
Co
rrect
Rate
[%
]
CLASSIFIER PERFORMANCE
'Rest' ->89.65%
'Move'->99.95%
'Total' ->95.10%
1.8 2 2.2 2.4 2.6
2
2.5
3
3.5
1st CSP - Log Features
2n
d C
SP
- L
og
Featu
res
0
1
Support Vectors
PREDICTING RESULTS
Analysis of the BCI
output calculated with
parameters extracted
from the same dataset
Plot of each trial in
the features
space
Haptics
Symposium
2014
Haptics
Symposium
2014
MODEL
EEG amp
CSP and LDA
weightsSpatial Filtering
Features ExtractionClassifier
User Interface
Haptics
Symposium
2014
Haptics
Symposium
2014
SHOWING RESULTs
[Frisoli et al. 2012]