Você está na página 1de 32

http://bci.ucsd.edu/Home.

html
Classification of the Event Related
EEG for the Purpose of Brain
Computer Interface (BCI)
Outlines
Introduction

Problem Formulation

Proposed Algorithm

Result

Conclusion

Demonstration

del R Milan, J.; Carmena, J.M., "Invasive or Noninvasive: Understanding Brain-Machine Interface Technology [Conversations in
BME]," Engineering in Medicine and Biology Magazine, IEEE , vol.29, no.1, pp.16,22, Jan.-Feb. 2010

Brain Computer Interface (BCI)
The BCI is a direct communication pathway between the brain
and the computer.


Rao, R. P N; Scherer, R., "Brain-Computer Interfacing [In the Spotlight]," Signal Processing Magazine, IEEE , vol.27, no.4,
pp.152,150, July 2010

Typical Block Diagram Of BCI
http://www.ese.wustl.edu/~nehorai/research/eegmeg/EMEG-Overview.html
http://blogs.oem.indiana.edu/scholarships/index.php/2009/10/26/neurons-and-electrodes/fmri_groot/
Signal Acquisition Process
Types Of BCI:
- Invasive (ECoG)
- Non-Invasive (i.e. EEG, fMRI)





Why EEG?
- Noninvasive
- Cost Effective
- Mobility




https://www.city.ac.uk/arts-social-sciences/psychology/research/cognitive-neuroscience-research-unit/eeg

Electroencaphalogram (EEG)










The Electroencephalogram (EEG) is the process of recording of
the electrical activity of the brain recorded from the scalp.

http://www.medicine.mcgill.ca/physio/vlab/biomed_signals/EEG_n.htm

About EEG
EEG Waves Classification:

- Delta (<3Hz)
- Theta (3.5Hz to 7.5Hz)
- Alpha (7.5Hz to 13Hz)
- Beta (>14Hz)


Detectable EEG Phenomena:

- Oscillatory Process
- Event Related Potential (ERP)
* P300





http://thebrain.mcgill.ca/
Human Nervous System
The Central Nervous System The Peripheral Nervous System
http://thebrain.mcgill.ca/
Spatial Classification Of Brain
Two Hemispheres
- Right
- Left

Four Main Lobes
- Frontal - Parietal
- Occipital - Temporal






http://thebrain.mcgill.ca/
Recording System











The International 10/20 System is a standard for recording EEG.

Motor Imagery (MI)
Motor Imagery (MI) task is the
imagination of the motor tasks
rather to do it in actual.



According to the Jeannerod (1994,
1995), brains response to conscious
Motor Imagery Task is functionally
equivalent to the brains response to
the preparation of the unconscious
Motor Execution (ME).




Note: There is a significant
difference between the visual
imagery (VI) and the motor
imagery (MI). Visual Imagery is the
visualization of the specific task
whereas the motor imagery is the
kinesthetic feelings of the specific
task.

Martin Lotze , Pedro Montoya , Michael Erb , Ernst Hlsmann , Herta Flor, Uwe Klose, Niels Birbaumer, Wolfgang Grodd, Activation of Cortical and Cerebellar
Motor Areas during Executed and Imagined Hand Movements: An fMRI Study, Journal of Cognitive Neuroscience September 1999, Vol. 11, No. 5, Pages 491-501

Spatial Reasoning Of The Brain
Regions For MI
Many studies have addressed the following spatial brain regions to be
activated during motor imagery task:
- SMA (Supplementary Motor Area)
- PMC (PreMotor Cortex)
- M1 (Primary Motor Cortex)

Solodkin Ana, Hlustik Petr, Chen, E. Elinor, Small, Steven L., Fine Modulation in Network Activation during Motor
Execution and Motor Imagery, Cereb Cortex. 2004 Nov;14(11):1246-55. Epub 2004 May 27.
Spatial Reasoning Of The Brain
Regions For MI










Mean Volume Of Activation (mm
3
) Per Region
http://thebrain.mcgill.ca/
Assumptions In Our Study
The left hemisphere is responsible for controlling the right
part of the body.

The source activation during the MI tasks happens mostly
in the sensorimotor areas.

Challenges In BCI
- Nonlinearity in EEG

- SNR

- Artefacts Rejection

- Computational Cost

Problem Formulation
- Motor Imagery (MI)

- Motor Execution (ME)

- Visual Imagery (VI)

- Any Stimulations in Human Body (i.e. Response to audio
signal )

http://www.bbci.de/competition/iii/desc_IVa.html
About The Dataset
Dataset IVa from the BCI Competition III:

-BrainAmp amplifiers

-Five Healthy Subjects

-118 Channels (Extended 10/20 System)

-Sampling Rate: 1 kHz ( Downsampled to 100 Hz for Analysis)

-Classes: left hand, right hand, right foot

-280 trials per subject

- Duration: 3.5s per trials (3s per trials after 0.5s of visual cues)

Block Diagram Of Proposed
Algorithm
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
Raw EEG
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
Butterworth Filter
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
Filter Order: 10 PassBand: 8Hz-40Hz

Common Spatial Pattern (CSP)
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
The CSP gives projection of a signal so that the difference
between the two classes is maximized.

Actually, the CSP projected matrix align the channels into a
space where a channel shows maximum variances for a class
while showing minimum variances for other class.

The projection matrix is estimated using sample based
covariance matrix from the training trials.



Common Spatial Pattern (CSP)
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
The CSP gives projection of a signal so that the difference
between the two classes is maximized.

Common Spatial Pattern (CSP)
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
We have selected the four optimal channels for two classes.
- Green: For class 1
- Majenta: For class 2

Regularized CSP (RCSP)
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
In RCSP algorithm, the projection matrix is estimated using the covariance
matrix which is regularized using different parameters.

The parameters are tuned to bias the covariance estimation and thus
sometimes gives better results where effect of outliers are significant.

The RCSP effectively increases the classification accuracies in small sample
settings.
Feature Extraction
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
Wavelet Decomposition
Shannon
Entropy
4
Two Layer Feed Forward Neural
Network
Raw EEG
Bandpass
Filtering
Spatial
Filtering
Feature
Extraction
Classification
- Training Phase : 120 Trials
- Validation Phase : 60 Trials
- Testing Phase : 100 Trials
Results
Result Analysis
- The performances are subject specific.

- Most of the cases, RCSP gives better result.

- For Subject 4 and Subject 5, we have got the best results in Study 4.

- For Subject 1 and Subject 2, the results are comparable.

- Only for Subject 3, we can see a little bit more difference between
Study 1 and Study 4.

- In case of average results, the best result is 88.78% and the result for
Study 4 is 87.5%.



Conclusion
As far, we have studied

Brain Computer Interface (BCI) System
Motor Imagery (MI) Task
Spatial Brain Dynamics
Proposed Algorithm
Result

The consideration of the event specific areas effectively
reduces the number of channels as well as the computational
cost [i.e. It is possible to select particular set of channels from
specific Region Of Interests (ROIs)].


References
+ Martin Lotze, Ulrike Halsband, Motor imagery, Journal of Physiology-Paris, Volume 99, Issues 46, June 2006, Pages
386-395, ISSN 0928-4257.
+ Martin Lotze , Pedro Montoya , Michael Erb , Ernst Hlsmann , Herta Flor, Uwe Klose, Niels Birbaumer, Wolfgang
Grodd, Activation of Cortical and Cerebellar Motor Areas during Executed and Imagined Hand Movements: An fMRI
Study, Journal of Cognitive Neuroscience September 1999, Vol. 11, No. 5, Pages 491-501
+ Solodkin Ana, Hlustik Petr, Chen, E. Elinor, Small, Steven L., Fine Modulation in Network Activation during Motor
Execution and Motor Imagery, Cereb Cortex. 2004 Nov;14(11):1246-55. Epub 2004 May 27.
+ Haiping Lu, How-Lung Eng, Cuntai Guan, K.N. Plataniotis, and A.N. Venetsanopoulos, "Regularized Common Spatial
Pattern With Aggregation for EEG Classification in Small-Sample Setting", IEEE Trans. on Biomedical Engineering, Vol. 57,
No. 12, pp. 2936-2946, Dec. 2010.
+ Lotte, F.; Cuntai Guan, "Regularizing Common Spatial Patterns to Improve BCI Designs: Unified Theory and New
Algorithms," Biomedical Engineering, IEEE Transactions on , vol.58, no.2, pp.355,362, Feb. 2011.
+ Yijun Wang; Shangkai Gao; Xiaorong Gao, "Common Spatial Pattern Method for Channel Selelction in Motor Imagery
Based Brain-computer Interface," Engineering in Medicine and Biology Society, 2005. IEEE-EMBS 2005. 27th Annual
International Conference of the , vol., no., pp.5392,5395, 17-18 Jan. 2006.
+ X. Yong, R. K. Ward, and G. Birch, "Robust common spatial patterns for EEG signal preprocessing," in EMBC 2008,
Vancouver, August 2008, pp. 20872090.
+ Blankertz, B.; Muller, K.; Curio, G.; Vaughan, T.M.; Schalk, G.; Wolpaw, J.R.; Schlogl, A.; Neuper, C.; Pfurtscheller, G.;
Hinterberger, T.; Schroder, M.; Birbaumer, N., "The BCI competition 2003: progress and perspectives in detection and
discrimination of EEG single trials," Biomedical Engineering, IEEE Transactions on , vol.51, no.6, pp.1044,1051, June 2004.
+ Ramoser, H.; Muller-Gerking, J.; Pfurtscheller, G., "Optimal spatial filtering of single trial EEG during imagined hand
movement," Rehabilitation Engineering, IEEE Transactions on , vol.8, no.4, pp.441,446, Dec 2000
+ Guido Dornhege, Benjamin Blankertz, Gabriel Curio, and Klaus-Robert Mller. Boosting bit rates in non-invasive EEG
single-trial classifications by feature combination and multi-class paradigms. IEEE Trans. Biomed. Eng., 51(6):993-1002,
June 2004.

Demo
..\Study Materials\Animation(s)\Mind Over
Mechanics_(1080p).mp4

..\Study Materials\Animation(s)\Paralyzed woman uses
thoughts to sip coffee from Brown University on Vimeo.mp4


Questions . . .?

Você também pode gostar