Anda di halaman 1dari 51

Brain Computer Interfacing Seminar Report 2011

SEMINAR REPORT ON
BRAIN COMPUTER INTERFACING
BY
ASHISH VISWANATH PRAKASH
DEPARTMENT OF ELECTRONICS AND
COMMUNICATION
COLLEGE OF ENGINEERING
CUSAT
1ULY 2011
1
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
ABSTRACT
A brain-machine interface is a communication sstem that !oes not !epen! on the brains normal
output path"as of peripheral ner#es an! muscles. It is a ne" communication lin$ bet"een a
functioning human brain an! the outsi!e "orl!. %hese are electronic interfaces "ith the brain&
"hich has the abilit to sen! an! recei#e signals from the brain. B'I uses brain acti#it to
comman!& control& actuate an! communicate "ith the "orl! !irectl through brain integration
"ith peripheral !e#ices an! sstems. %he signals from the brain are ta$en to the computer #ia the
implants for !ata entr "ithout an !irect brain inter#ention. B'I transforms mental !ecisions
an!(or reactions into control signals b anal)ing the bioelectrical brain acti#it.
*hile lin$ing the brain !irectl "ith machines "as once consi!ere! science fiction& a!#ances
o#er the past fe" ears ha#e ma!e it increasingl #iable. It is an area of intense research "ith
almost limitless possibilities. %he human brain is the most comple+ phsical sstem "e $no" of&
an! "e "oul! ha#e to un!erstan! its operation in great !etail to buil! such a !e#ice. An
imme!iate goal of brain-machine interface stu! is to pro#i!e a "a for people "ith !amage!
sensor(motor functions to use their brain to control artificial !e#ices an! restore lost
capabilities. B combining the latest !e#elopments in computer technolog an! hi-tech
engineering& paral)e! persons "ill be able to control a motori)e! "heel chair& computer painter&
or robotic arm b thought alone. In this era "here !rastic !iseases are getting common it is a
boon if "e can !e#elop it to its full potential. Recent technical an! theoretical a!#ances& ha#e
!emonstrate! the ultimate feasibilit of this concept for a "i!e range of space-base!
applications. Besi!es the clinical purposes such an interface "oul! fin! imme!iate applications
in #arious technolog pro!ucts also.
,e"or!s- Brain&.euro-electronics&Electro encephalograph& interfacing&
2
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
INTRODUCTION
/icture a time "hen humans see in the 01 an! IR portions of the electromagnetic spectrum& or
hear speech on the nois flight !ec$ of an aircraft carrier2 or "hen sol!iers communicate b
thought alone. Imagine a time "hen the human brain has its o"n "ireless mo!em so that instea!
of acting on thoughts& "ar fighters ha#e thoughts that act. Imagine that one !a "e "ill be able
to !o"nloa! #ast amounts of $no"le!ge !irectl to our brain3 So as to cut the length processes
of learning e#erthing from scratch. Instea! of paing to go to uni#ersit "e coul! pa to get a
4$no"le!ge implant4 an! perhaps be able to obtain man lifetimes "orth of $no"le!ge an!
e+pertise in #arious fiel!s at a oung age In the near future& most !e#ices "oul! be
remote(logicall controlle!. Researchers are close to brea$throughs in neural interfaces& meaning
"e coul! soon mesh our min!s "ith machines. %his technolog has the capabilit to impact our
li#es in "as that ha#e been pre#iousl thought possible in onl sci-fi mo#ies.
%here is legitimate scientific interest in the possibilit of connectingbrains an! computers5from
pro!ucing robotic limbs controlle!!irectl b brain acti#it to altering memor an! moo! "ith
implante!electro!es to the far-out prospect of becoming immortal b 6uploa!ing7our min!s into
machines. %his area of in8uir has seen remar$ablea!#ances in recent ears& man of them
aime! at helping the se#erel !isable!to replace lost functions. 9et public un!erstan!ing of this
research isshape! b sensationalistic an! mislea!ing co#erage in the press2 it is colore!b
!eca!es of fantastical science fiction portraals2 an! it is !istorte!b the utopian hopes of a small
but #ocal ban! of enthusiasts "ho !esire to eliminate the boun!aries bet"een brains an!
machines as part of a larger 6transhumanist7 pro:ect
Neuroelectronics
.euroelectronics& sometimes referre! to as neurotechnolog& is the !iscipline that !eals "ith the
interface bet"een the human ner#ous sstem an! electronic !e#ices. It is a highl comple+ an!
inter!isciplinar fiel! "ith contributions from computer science& cogniti#e science& neurosurger
an! biome!ical engineering. .euroelectronics has roughl three relate! branches- ;1<
neuroimaging& ;2< brain-computer interfaces ;BCIs<& an! ;=< electrical neural stimulation. %he
!iscipline e+ists for more than half a centur. >o"e#er& in the last !eca!e significant a!#ances
=
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
ha#e been ma!e& particularl in neuroimaging& "hich re#olutioni)e! the fiel! b allo"ing
researchers to !irectl monitor brain acti#it !uring e+periments. An! it is pre!icte! that
neuroelectronics& particularl neuroimaging an! brain-computer interfacing& "ill be emploe!
much more in the future.
BCIs& sometimes calle! brain-machine interfaces ;B'Is<& are an emerging neurotechnolog that
translates brain acti#it into comman! signals for e+ternal !e#ices. Research on BCIs began in
the 1?@0s at the 0ni#ersit of California Aos Angeles ;0CAA<. Researchers at 0CAA also
coine! the term brain-computer interface. A BCI establishes a !irect communication path"a
bet"een the brain an! the !e#ice to be controlle!. %he are mainl being !e#elope! for me!ical
reasons& because there is a societal !eman! for technologies "hich help to restore functions of
humans "ith central ner#ous sstem ;C.S<. /atients for "hom a BCI "oul! be useful usuall
ha#e !isabilities in motor function or communication. %his coul! be ;partl< restore! b using a
BCI to steer a motori)e! "heelchair& prosthesis& or b selecting letters on a computer screen "ith
a cursor. In#asi#e or non-in#asi#e electro!es are use! to !etect brain acti#it& "hich is
subse8uentl translate! b a signal processing unit into comman! signals for the e+ternal !e#ice.
%he most common BCI respon!s to specific patterns !etecte! in spatiotemporal EEBs measure!
non-in#asi#el from the scalp. Spatiotemporal EEBs can be controlle! b imagining specific
mo#ements ;Basson C *ar"ic$& 200@<. So& merel b imagining mo#ements one can steer a
"heelchair& prosthesis or a cursor on a computer screen.
%here are roughl three branches in neuroelectronics. Each branch uses !ifferent !e#ices to
interface "ith the brain& an! each of these !e#ices has !ifferent features. %he first branch&
neuroimaging& uses techni8ues such as f'RI& /E%& 'EB or EEB& amongst others& to e+tract
information from the brain to !iagnose !isor!ers or to stu! the brain. %he secon! branch& BCIs&
uses in#asi#e or non-in#asi#e electro!es to e+tract information from the brain& not for !iagnostic
or research purposes& but to control e+ternal !e#ices such as "heelchairs& computers or airplanes.
An! the thir! branch& electrical neural stimulation& uses in#asi#e electro!es to sen! electrical
signals to specific parts of the brain. %he onl !efining feature these three branches ha#e in
common is that the all interface electrical !e#ices "ith the brain& either to e+tract information
from the brain or to sen! electrical signals to the brain.
D
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
HISTORY
%he brain has been clearl un!erstoo! to be the seat of the min! for less than four centuries. A
number of anatomists& philosophers& an! phsiciansha!& since the !as of the ancient Bree$s&
conclu!e! that the soul "as resi!ent in the hea!. In 1?E0& at the height of interest in cbernetics&
the "or! cborg5short for 6cbernetic organism75"as coine! b researcher 'anfre! E.Clnes
in a paper he co-"rote for the :ournal Astronautics Aroun! the same time& Fac$ E. Steele&a
polmath !octor-engineer-neuroanatomist ser#ing in the 0.S. Air Gorce&
coine! the "or! bionics for the use of principles !eri#e! from li#ing sstems to sol#e
engineering an! !esign problems. But to the most ambitious an! most ra!ical a!#ocates of
merging brains an! machines& such a!#ances are mere chil!Hs pla. %hese so-calle!
transhumanists long for an age "hen human beings "ill lea#e the miser ies an! limits of the
bo! behin!& an! achie#e ne" ecstasies of free!om
an! control- *e "ill sen! feelings an! conscious thoughts !irectl frombrain to brain2 "e "ill
!elete un"ante! memories at "ill2 "e "ill master
comple+ sub:ects b 6!o"nloa!ing7 them !irectl into our min!s2 "e "ill6:ac$ in7 to #irtual
realities2 an! e#entuall& "e "ill be able to 6uploa!7
our personalities into computers or robots& "here the self coul! li#e on in!efinitel.
%he histor of Brain-Computer-Interfaces ;BCI< starts "ith >ans BergerIs!isco#er of the
electrical acti#it of human brain an! the !e#elopment ofelectroencephalograp ;EEB<.Berger
stu!ie! me!icine at the 0ni#ersit of Fena an! recei#e! his !octorate in
1J?@. >e became a professor in 1?0E an! the !irector of the 0ni#ersitIspschiatr an!
neurolog clinic in 1?1?. In 1?2D Berger "as the first one "horecor!e! an EEB from a human
brain. B anal)ing EEBs Berger "as able toi!entif !ifferent "a#es or rhthms "hich are
present in a brain& as the Alpha *a#e ;J K 12 >)<&also $no"n as BergerIs *a#e.
In the 1?J0s& Apostolos Beorgopoulos at Fohns >op$ins 0ni#ersit foun! a mathematical
relationship bet"een the electrical responses of single motor-corte+ neurons in rhesus maca8ue
mon$es an! the !irection that mon$es mo#e! their arms ;base! on a cosine function<. >e also
foun! that !isperse! groups of neurons in !ifferent areas of the brain collecti#el controlle!
L
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
motor comman!s but "as onl able to recor! the firings of neurons in one area at a time because
of technical limitations impose! b his e8uipment
Prominent research successes
In 1???& researchers le! b Barrett Stanle at >ar#ar! 0ni#ersit !eco!e! neuronal firings to
repro!uce images seen b cats. %he team use! an arra of electro!es embe!!e! in the thalamus
;"hich integrates all of the brainHs sensor input< of sharp-ee! cats. Researchers targete! 1@@
brain cells in the thalamus lateral geniculate nucleus area& "hich !eco!es signals from the retina.
%he cats "ere sho"n eight short mo#ies& an! their neuron firings "ere recor!e!. 0sing
mathematical filters& the researchers !eco!e! the signals to generate mo#ies of "hat the cats sa"
an! "ere able to reconstruct recognisable scenes an! mo#ing ob:ects.
'iguel .icolelis has been a prominent proponent of using multiple electro!es sprea! o#er a
greater area of the brain to obtain neuronal signals to !ri#e a BCI. Such neural ensembles are
sai! to re!uce the #ariabilit in output pro!uce! b single electro!es& "hich coul! ma$e it
!ifficult to operate a BCI
B 2000& the group succee!e! in buil!ing a BCI that repro!uce! o"l mon$e mo#ements "hile
the mon$e operate! a :ostic$ or reache! for foo!.%he BCI operate! in real time an! coul! also
control a separate robot remotel o#er Internet protocol. But the mon$es coul! not see the arm
mo#ing an! !i! not recei#e an fee!bac$& a so-calle! open-loop BCI Mther labs that !e#elop
BCIs an! algorithms that !eco!e neuron signals inclu!e Fohn Donoghue from Bro"n 0ni#ersit&
An!re" Sch"art) from the 0ni#ersit of /ittsburgh an! Richar! An!ersen from Caltech
DonoghueIs group reporte! training rhesus mon$es to use a BCI to trac$ #isual targets on a
computer screen "ith or "ithout assistance of a :ostic$ ;close!-loop BCI<.Sch"art)Is group
E
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
create! a BCI for three-!imensional trac$ing in #irtual realit an! also repro!uce! BCI control in
a robotic arm
Cell-culture BCIs
Research into techni8ues for stimulating an! recor!ing from in!i#i!ual neurons gro"n on
semicon!uctor chips is sometimes referre! to as neuroelectronics or neurochips.
*orl! first- .eurochip !e#elope! b Caltech researchers Ferome /ine an! 'ichael 'aher
De#elopment of the first "or$ing neurochip "as claime! b a Caltech team le! b Ferome /ine
an! 'ichael 'aher in 1??@. %he Caltech chip ha! room for 1E neurons In 200=& a team le! b
%heo!ore Berger at the 0ni#ersit of Southern California starte! "or$ on a neurochip !esigne!
to function as an artificial or prosthetic hippocampus. %he neurochip "as !esigne! to function in
rat brains an! is inten!e! as a prototpe for the e#entual !e#elopment of higher-brain prosthesis
%homas De'arse at the 0ni#ersit of Glori!a use! a culture of 2L&000 neurons ta$en from a ratIs
brain to fl a G-22 fighter :et aircraft simulator.After collection& the cortical neurons "ere
culture! in a petri !ish an! rapi!l begin to reconnect themsel#es to form a li#ing neural
net"or$. %he cells "ere arrange! o#er a gri! of E0 electro!es an! traine! to control the pitch an!
a" functions of the simulator
READING THE BRAIN - NEUROIMAGING IN BCIS
Neuroimaging definition
6.euroimaging inclu!es the use of #arious techni8ues to either !irectl or in!irectl image
thestructure& function& or pharmacolog of the brain. It is a relati#el ne" !iscipline "ithin
me!icinean! neuroscience.7 N*I,IO.IP
Direct Neural Contact
%his is the most accurate metho! of recor!ing potentials occuring in the brain as it has
!irectcontact to e#er neuron in the brain& e.g. b using nanorobots. .ee!less to sa that this
metho! ishighl in#asi#e an! impracticable "ith respect to our current technolog. >o"e#er&
"ith ongoinga!#ances in nanotechnolog this metho! might become realit.
Electroencephalography (EEG)
%his proce!ure is the first non-in#asi#e neuroimaging techni8ue !isco#ere!. It measures
theelectrical acti#it of the brain. Due to its ease of use& cost an! high temporal resolution this
metho!is the most "i!el use! one in BCIs no"a!as. >o"e#er& in practice EEBs are highl
susceptible tonoise an! thus re8uire a significant amount of user training in or!er to be operable
@
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
in a BCI.Auc$il& recent research at the Graunhofer Societ N*I,IOBCIP has sho"n that this
problem can beo#ercome b using neural net"or$s to shift the learning o#erhea! from the
human to the computer.
Magnetoencephalography (MEG)
%hough similar to the EEB in that it is a non-e#asi#e technolog the 'EB is a much ne"er
an!more accurate technolog. Instea! of measuring the electrical acti#it in the brain this
technologrecor!s magnetic fiel!s pro!uce! b it. %he main !ra"bac$s of this technolog are its
highere8uirements in e8uipment.
60sing 'EB re8uires a room fille! "ith super-con!ucting magnets an! giant super-cooling
heliumtan$s surroun!e! b shiel!e! "alls.7
Functional Magnetic Resonance Imaging (fMRI)
%his techni8ue measures the haemo!namic response ;bloo! flo" an! bloo! o+genation<
relate! toneural acti#it in the brain b the use of 'RI ;'agnetic Resonance Imaging formerl
$no"n as'agnetic Resonance %omograph 'R%<. %he fact that there is a correlation bet"een
neural acti#itan! the brainHs haemo!namics ma$es the f'RI a neuroimaging tool. In contrast
to the 'RI "hichstu!ies the brainHs structure this metho! stu!ies the brainHs function. As this
metho! re8uires 'RItechnolog it nee!s #er special e8uipment an! thus is 8uite costl.
BLOCK DIAGRAM
J
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
?
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
BLOCK DESCRIPTION
%he B'I consists of se#eral components- 1.the implant !e#ice& or chronic multi-electro!e arra&
2.the signal recor!ing an! processing section& =.an e+ternal !e#ice the sub:ect uses to pro!uce
an! control motion an! D.a fee!bac$ section to the sub:ect. %he first component is an implante!
arra of microelectro!es into the frontal an! parietal lobes5areas of the brain in#ol#e! in
pro!ucing multiple output comman!s to control comple+ muscle mo#ements. %his !e#ice recor!
action potentials of in!i#i!ual neurons an! then represent the neural signal using a rate co!e .%he
secon! component consists of spi$e !etection algorithms& neural enco!ing an! !eco!ing
sstems& !ata ac8uisition an! real time processing sstems etc .A high performance DS/
architecture is use! for this purpose. %he e+ternal !e#ice that the sub:ect uses ma be a robotic
arm& a "heel chair etc. !epen!ing upon the application. Gee!bac$ is an important factor in
BCIQs. In the BCIQs base! on the operant con!itioning approach& fee!bac$ training is essential
for the user to ac8uire the control of his or her EEB response. >o"e#er& fee!bac$ can spee! up
the learning process an! impro#e performance.
the three tpes of BCI. %hese tpes are !eci!e! on the basis of the techni8ue use! for the
interface. Each of these techni8ues has some a!#antages as "ell as some !isa!#antages.
Invasive BCI:
In#asi#e BCI are !irectl implante! into the gre matter of the brain !uring neurosurger. %he
pro!uce the highest 8ualit signals of BCI !e#ices . In#asi#e BCIs has targete! repairing
!amage! sight an! pro#i!ing ne" functionalit to paral)e! people. But these BCIs are prone to
buil!ing up of scar-tissue "hich causes the signal to become "ea$er an! e#en lost as bo! reacts
to a foreign ob:ect in the brain
Partially Invasive BCI:
/artiall in#asi#e BCI !e#ices are implante! insi!e the s$ull but rest outsi!e the brain rather than
ami!st the gre matter. %he pro!uce better resolution signals than non-in#asi#e BCIs "here the
bone tissue of the cranium !eflects an! !eforms signals an! ha#e a lo"er ris$ of forming scar-
tissue in the brain than full-in#asi#e BCIs.
Electrocorticograph;ECoB< uses the same technolog as non-in#asi#e electroencephalograph&
but the electro!es are embe!!e! in a thin plastic pa! that is place! abo#e the corte+& beneath the
!ura mater. ECoB technologies "ere first trale! in humans in 200D b Eric Aeuthar!t an! Daniel
'oran from *ashington 0ni#ersit in St Aouis. In a later trial& the researchers enable! a teenage
bo to pla Space In#a!ers using his ECoB implant. %his research in!icates that it is !ifficult to
pro!uce $inematic BCI !e#ices "ith more than one !imension of control using ECoB.
10
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
Non-Invasive BCI :
As "ell as in#asi#e e+periments& there ha#e also been e+periments in humans using non-in#asi#e
neuroimaging technologies as interfaces. Signals recor!e! in this "a ha#e been use! to po"er
muscle implants an! restore partial mo#ement in an e+perimental #olunteer. Although the are
eas to "ear& non-in#asi#e implants pro!uce poor signal resolution because the s$ull !ampens
signals& !ispersing an! blurring the electromagnetic "a#es create! b the neurons. Although the
"a#es can still be !etecte! it is more !ifficult to !etermine the area of the brain that create! them
or the actions of in!i#i!ual neurons
MAIN PRINCIPLE
'ain principle behin! this interface is the bioelectrical acti#it of ner#es an! muscles. It is no"
"ell establishe! that the human bo!& "hich is compose! of li#ing tissues& can be consi!ere! as
a po"er station generating multiple electrical signals "ith t"o internal sources& namel muscles
an! ner#es.
*e $no" that brain is the most important part of human bo!. It controls all the emotions an!
functions of the human bo!. %he brain is compose! of millions of neurons. %hese neurons "or$
together in comple+ logic an! pro!uce thought an! signals that control our bo!ies. *hen the
neuron fires& or acti#ates& there is a #oltage change across the cell& ;R100m#< "hich can be rea!
through a #ariet of !e#ices. *hen "e "ant to ma$e a #oluntar action& the comman! generates
from the frontal lobe. Signals are generate! on the surface of the brain. %hese electric signals are
!ifferent in magnitu!e an! fre8uenc.
B monitoring an! anal)ing these signals "e can un!erstan! the "or$ing of brain. *hen "e
imagine oursel#es !oing something& small signals generate from !ifferent areas of the brain.
%hese signals are not large enough to tra#el !o"n the spine an! cause actual mo#ement. %hese
small signals are& ho"e#er& measurable. A neuron !epolari)es to generate an impulse2 this action
causes small changes in the electric fiel! aroun! the neuron. %hese changes are measure! as 0
;no impulse< or 1 ;impulse generate!< b the electro!es. *e can control the brain functions b
artificiall pro!ucing these signals an! sen!ing them to respecti#e parts. %his is through
stimulation of that part of the brain& "hich is responsible for a particular function using
implante! electro!es.
BMI COMPONENTS
11
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
A brain-machine interface ;B'I< in its scientific interpretation is a combination of se#eral
har!"are an! soft"are components tring to enable its user to communicate "ith a computer b
intentionall altering his or her brain "a#es. %he tas$ of the har!"are part is to recor! the
brain"a#esK in the form of the EEB signal K of a human sub:ect& an! the soft"are has to anal)e
that !ata. In other "or!s& the har!"are consists of an EEB machine an! a number of electro!es
scattere! o#er the sub:ectQs s$ull. %he EEB machine& "hich is connecte! to the electro!es #ia
thin "ires& recor!s the brain-electrical acti#it of the sub:ect& iel!ing a multi-!imensional
;analog or !igital< output. %he #alues in each !imension ;also calle! channel< represent the
relati#e !ifferences in the #oltage potential measure! at t"o electro!e sites.
%he soft"are sstem has to rea!& !igiti)e ;in the case of an analog EEB machine<& an!
preprocess the EEB !ata ;separatel for each channel<& Sun!erstan! the sub:ectQs intentions&
an! generate appropriate output. %o interpret the !ata& the stream of EEB #alues is cut into
successi#e segments& transforme! into a stan!ar!i)e! representation& an! processe! "ith the help
of a classifier. %here are se#eral !ifferent possibilities for the reali)ation of a classifier2 one
approach K in#ol#ing the use of an artificial neural net"or$ ;A..< K has become the metho! of
choice in recent ears.
. IMPLANT DEVICE
%he EEB is recor!e! "ith electro!es& "hich are place! on the scalp. Electro!es are small plates&
"hich con!uct electricit. %he pro#i!e the electrical contact bet"een the s$in an! the EEB
12
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
recor!ing apparatus b transforming the ionic current on the s$in to the electrical current in the
"ires. %o impro#e the stabilit of the signal& the outer laer of the s$in calle! stratum corneum
shoul! be at least partl remo#e! un!er the electro!e. Electrolte gel is applie! bet"een the
electro!e an! the s$in in or!er to pro#i!e goo! electrical contact.
Gigure D.An arra of microelectro!es
0suall small metal-plate electro!es are use! in the EEB recor!ing. .eural implants can be use!
to regulate electric signals in the brain an! restore it to e8uilibrium. %he implants must be
monitore! closel because there is a potential for almost anthing "hen intro!ucing foreign
signals into the brain.
%here are a fe" ma:or problems that must be a!!resse! "hen !e#eloping neural implants. %hese
must be ma!e out of biocompatible material or insulate! "ith biocompatible material that the
bo! "onQt re:ect an! isolate. %he must be able to mo#e insi!e the s$ull "ith the brain "ithout
causing an !amage to the brain. %he implant must be chemicall inert so that it !oesnQt interact
"ith the hostile en#ironment insi!e the human bo!. All these factors must be a!!resse! in the
case of neural implants2 other"ise it "ill stop sen!ing useful information after a short perio! of
time.
%here are simple single "ire electro!es "ith a number of !ifferent coatings to comple+ three-
!imensional arras of electro!es& "hich are encase! in insulating biomaterials. Implant re:ection
an! isolation is a problem that is being a!!resse! b !e#eloping biocompatible materials to coat
or incase the implant.
Mne option among the biocompatible materials is %eflon coating that protects the implant from
the bo!. Another option is a cell resistant snthetic polmer li$e pol#inl alcohol. %o $eep the
implant from mo#ing in the brain it is necessar to ha#e a fle+ible electro!e that "ill mo#e "ith
the brain insi!e the s$ull. %his can ma$e it !ifficult to implant the electro!e. Dipping the micro
!e#ice in polethlene glcol& "hich causes the !e#ice to become less fle+ible& can sol#e this
problem
1=
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
problem. Mnce in contact "ith the tissue this coating 8uic$l !issol#es. %his allo"s eas
implantation of a #er fle+ible implant.
%hree-!imensional arras of electro!es are also un!er !e#elopment. %hese !e#ices are
constructe! as t"o-!imensional sheet an! then bent to form =D arra. %hese can be constructe!
using a polmer substrate that is then fitte! "ith metal lea!s. %he are !ifficult to implement& but
gi#e a much great range of stimulation or sensing than simple ones.
Gigure L. Bloc$ !iagram of the neurotrophic electro!es for implantation in human patients
.
A microscopic glass cone contains a neurotrophic factor that in!uces neurites to gro" into the
cone& "here the contact one of se#eral gol! recor!ing "ires. .eurites that are in!uce! to gro"
into the glass cone ma$e highl stable contacts "ith recor!ing "ires. Signal con!itioning an!
telemetric electronics are full implante! un!er the s$in of the scalp. An implante! transmitter
;%T< sen!s signals to an e+ternal recei#er ;RT<& "hich is connecte! to a computer.
SIGNAL PROCESSING SECTION
Multichannel Acquisition Systems
Electro!es interface !irectl to the non-in#erting opamp inputs on each channel. At this section
amplification& initial filtering of EEB signal an! possible artifact remo#al ta$es place. Also A(D
con#ersion is ma!e& i.e. the analog EEB signal is !igiti)e!. %he #oltage gain impro#es the
signal-to-noise ratio ;S.R< b re!ucing the rele#ance of electrical noise incurre! in later stages.
/rocesse! signals are time-!i#ision multiple+e! an! sample!.
1D
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
SPIKE DETECTION
Real time spi$e !etection is an important re8uirement for !e#eloping brain machine interfaces.
Incorporating spi$e !etection "ill allo" the B'I to transmit onl the action potential "a#eforms
an! their respecti#e arri#al times instea! of the sparse& ra" signal in its entiret. %his
compression re!uces the transmitte! !ata rate per channel& thus increasing the number of
channels that ma be monitore! simultaneousl. Spi$e !etection can further re!uce the !ata rate
if spi$e counts are transmitte! instea! of spi$e "a#eforms. Spi$e !etection "ill also be a
necessar first step for an future har!"are implementation of an autonomous spi$e sorter.
Gigure E sho"s its implementation using an application-specific integrate! circuit ;ASIC< "ith
limite! computational resources. A lo" po"er implantable ASIC for !etecting an! transmitting
neural spi$es "ill be an important buil!ing bloc$ for B'Is. A har!"are reali)ation of a spi$e
!etector in a "ireless B'I must operate in real-time& be full autonomous& an! function at
realistic signal-to- noise ratios ;S.Rs<.
An implante! ASIC con!itions signal from e+tra cellular neural electro!es& !igiti)es them& an!
then !etects A/ spi$es. %he spi$e "a#eforms are transmitte! across the s$in to a B'I processor&
"hich sorts the spi$es an! then generates the comman! signals for the prosthesis.
Classical metho!s for remo#ing eeblin$ artifacts can be classifie! into re:ection
metho!s an! subtraction metho!s ND.EP-
U Rejection methods consist of !iscar!ing contaminate! EEB& base! on eitherautomatic or
#isual !etection. %heir success cruciall !epen!s on the 8ualit of the
1L
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
!etection& an! its use !epen!s also on the specific application for "hich it is use!.%hus& although
for epileptic applications& it can lea! to an unacceptable loss of !ata&for others& li$e a Brain
Computer interface& its use can be a!e8uate.
U Subtraction methods are base! on the assumption that the measure! EEB is a
linearcombination of an original EEB an! a signal cause! b ee mo#ement& calle!
EMB;electrooculogram<. %he EMB is a potential pro!uce! b mo#ement of the ee oreeli!
;Gig. D-D<. %he original EEB is hence reco#ere! b subtracting separatelrecor!e! EMB from the
measure! EEB& using appropriate "eights ;re:ecting theinfluence of the EMB on particular EEB
channels<.
Artifact rejection based on peak elimination.
As pre#ious "or$s ha#e sho"n ND.=P& the presence of artifacts in EEB signalspro!uces a rapi!
increase of energ in forehea! locations Gp1 an! Gp2. %he metho!
!e#elope! here consists of the analsis of these t"o channels b small o#erlapping"in!o"s& in
or!er to chec$ if the energ of the signals surpasses an establishe! blin$threshol!. In case it !oes
it& the samples coming from the corrupte! signal are re:ecte!from all the EEB signals.Despite the
simplicit of this metho!& the results obtaine! ha#e been satisfactor enough to consi!er it as an
initial option for a real-time Brain Computer Interface. It iseas enough to be implemente! on a
lo" comple+it signal processing platform..e#ertheless& this metho! has the incon#enience of
re:ecting some non-corrupte! !ata inother scalp channels& as "ell as in the frontal channels.*ith
the purpose of impro#ing the !ata preser#ation& "e ha#e !e#elope! a similarsstem base! on a
8ua!ratic %ime Gre8uenc Representation of the signal. %his analsista$es a!#antage of the high
resolution of this techni8ue in time an! fre8uenc& for
establishing& after an appropriate training& the !ifferences bet"een a corrupte! EEBsignal an! a
non-corrupte! EEB signal& in or!er to be able of !istinguish an! re:ect theeeblin$ artifact. %his
techni8ue& "hich is base! on energ !istributions& is an alternati#eto the classical artifact
re:ection metho! presente! here& an! it shoul! be consi!ere! in afuture "or$.
Blinking artifact recognition using artificial neural network.
%he metho! propose! b Bogac) an! colleagues& use! a neural base! approach tofin! artifacts in
EEB signals ND.DP. %he input to the neural net"or$ "as not a ra" sample!signal& but !ifferent
coefficients compute! for a "in!o" of one secon! of the signal&
e+pressing some characteristic properties of blin$ing artifacts. D1 coefficients "ere!esigne!.
Some of them "ere !esigne! b the authors an! "ere base! on their $no"le!geabout the artifact
recognition& an! a total of 1D "ere chosen b terms of sensiti#it an!correlation. A large training
1E
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
set inclu!ing coefficients for o#er 2@000 "in!o"s "as use!&containing !ifferent $in!s of
blin$ing artifacts& pathological an! proper "a#es& an!artifacts cause! b other sources ;e.g. :a"&
muscle<. After"ar!s& three classificationalgorithms "ere teste! an! compare!- $-neighbors& RBG
net"or$s an! bac$ propagationnet"or$s. %he lo"est classification error ;1.D0V< "as obtaine!
for the bac$ propagationnet"or$& "ith a classification time of the test set ;E22@ "in!o"s< of 2
secon!s ND.DP.%his metho! achie#es high classification accurac than$s to t"o factors-
- Aarge training set containing !ifferent $in!s of EEB "a#es.
- %he coefficients !eli#ere! to the net"or$Hs inputs& e+press the characteristic
features of artifacts& since the enco!e large amount of !omain e+pertHs
$no"le!ge.
0nfortunatel& the first factor can be problematic in the use of a Brain computer
Interface.
4. 4. Artifact rejection based in bandpass FIR filters.
%he metho! propose! b Bupta an! colleagues use! a fi+e! ban!pass GIR filter&follo"e! b a
sub:ect specific eeblin$ threshol!& in or!er to remo#e the eeblin$ an!eeball mo#ement
artifacts. %his techni8ue& "hose bloc$ !iagram can be seen in Gig. D-L&consist of ND.1P-
1. /ass the ra" EEB samples obtaine! from analog-to-!igital con#erter through a
!igital ban!pass filter ;B/G< to remo#e slo" baseline !rift.
2. Determine the blin$ threshol! ;1t< for specific sub:ect in brief training session.
=. Compare the absolute sample #alue "ith 1t.
D. If the #alue is e+cee!e! then remo#e . samples from the #icinit of )ero crossing
;.(2 on either si!e of treshol! crossing<.
L. Shift the follo"ing . samples to fill up the gap create! b blin$ remo#al. %hese
gaps "ill& other"ise& grossl !istort the spectrum.
Figure 4-5. Scheme of the propose! sstem.
1@
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
%he e+periments carrie! out through this scheme& for Gp1 an! Gp2 electro!es
location& pro#i!e! interesting results in eeblin$ artifact re:ection. %his metho! presents
the a!#antage of "or$ing e#en un!er baseline !rift artifacts con!itions& an! also is eas
enough to be implemente! on a lo"-cost !igital signal processor& on a real K time sstem.
.e#ertheless& it fails if the blin$ rate is unnaturall high& an! the training session for each
in!i#i!ual is 8uite long- =0 sec ;E eeblin$s on a a#erage for a normal sub:ect<.
SIGNAL ANALYSIS
Geature e+traction an! classification of EEB are !ealt in this section. In this stage& certain
features are e+tracte! from the preprocesse! an! !igiti)e! EEB signal. In the simplest form a
certain fre8uenc range is selecte! an! the amplitu!e relati#e to some reference le#el measure!
%picall the features are fre8uenc content of the EEB signal< can be calculate! using& for
e+ample& Gast Gourier %ransform ;GG% function<. .o matter "hat features are use!& the goal is to
form !istinct set of features for each mental tas$. If the feature sets representing mental tas$s
o#erlap each other too much& it is #er !ifficult to classif mental tas$s& no matter ho" goo! a
classifier is use!. Mn the other han!& if the feature sets are !istinct enough& an classifier can
classif them. %he features e+tracte! in the pre#ious stage are the input for the classifier.
%he classifier can be anthing from a simple linear mo!el to a comple+ nonlinear neural net"or$
that can be traine! to recogni)e !ifferent mental tas$s. .o"a!as real time processing is use!
"i!el. Real-time applications pro#i!e an action or an ans"er to an e+ternal e#ent in a timel
an! pre!ictable manner. So b using this tpe of sstem "e can get output nearl at the same
time it recei#es input. %elemetr is han!le! b a "earable computer. %he host station accepts the
!ata #ia either a "ireless access point or its o"n !e!icate! ra!io car!.
EEG Signal Pre - Processing.
Mne of the main problems in the automate! EEB analsis is the !etection of the!ifferent $in!s of
interference "a#eforms ;artifacts< a!!e! to the EEB signal !uring therecor!ing sessions. %hese
interference "a#eforms& the artifacts& are an recor!e! electrical potentials not originate! in
brain. %here are four main sources of artifacts emission-
1. EEB e8uipment.
2. Electrical interference e+ternal to the sub:ect an! recor!ing sstem.
1J
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
=. %he lea!s an! the electro!es.
D. %he sub:ect her(himself- normal electrical acti#it from the heart& eeblin$ing& ees
mo#ement& an! muscles in general.
In case of #isual inspections& the artifacts can be 8uite easil !etecte! b EEBe+perts. >o"e#er&
!uring the automate! analsis these signal patterns often cause seriousmisclassifications thus
re!ucing the clinical usabilit of the automate! anal)ing sstems.Recognition an! elimination
of the artifacts in real K time EEB recor!ings is a comple+tas$& but essential to the !e#elopment
of practical sstems./re#ious "or$s ha#e sho"n that the most se#ere of the artifacts are !ue to
eeblin$s an! eeball mo#ements. A mo#ement of the eeball an! the eeli!s causes achange in
the potential fiel! because of the e+isting potential !ifference of about 100m1bet"een the
cornea an! the retina ND.1P. %his change affects mainl the signals from themost frontal
electro!es ;Gp1 an! Gp2 an! also other frontal electro!es- G=& GD& G@ an!GJ<& an! in!uces in
them man high an! lo" fre8uencies& !epen!ing upon its !urationan! amplitu!e
EEG SIGNAL CLASSIFICATION.
Mscillator states are the most remar$able features of EEB acti#it& because the
reflect not onl the snchroni)ation of massi#e numbers of neurons but also a temporall
or!ere! rhthmicit of acti#ation NL.EP. Different oscillator patterns ma be in!icati#e of
!ifferent information processing states& an! it has been propose! that the oscillator
patterns pla an acti#e role in these states NL.EP& NL.@P. Accor!ing to this #ie"& the
rhthmic snchroni)ation !uring oscillator states can ser#e to enhance perception&
learning& an! the transmission of neuronal signals bet"een !ifferent regions of the brain
Classifier Design:
%he !esign of the classifier is hea#il base! on Christin SchWferIs !esign use! for the Dataset III
of the BCI Competition II. Instea! of using a Baussian multi#ariate Baesian classifier& here "e
use a neural net to obtain the classification for each time instant t. %hose outputs are then
integrate! in time using a "eighte! sum. %he i!ea is simple- outputs "ith lo" confusion shoul!
ha#e higher "eights
EXTERNAL DEVICE
%he classifierQs output is the input for the !e#ice control. %he !e#ice control simpl transforms
the classification to a particular action. %he action can be& e.g.& an up or !o"n mo#ement of a
cursor on the fee!bac$ screen or a selection of a letter in a "riting application. >o"e#er& if the
classification "as Snothing or Sre:ect & no action is performe!& although the user ma be
1?
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
informe! about the re:ection. It is the !e#ice that sub:ect pro!uce an! control motion. E+amples
are robotic arm& thought controlle! "heel chair etc
FEEDBACK
Real-time fee!bac$ can !ramaticall impro#e the performance of a brainKmachine interface.
Gee!bac$ is nee!e! for learning an! for control. Real-time fee!bac$ can !ramaticall impro#e
the performance of a brainKmachine interface. In the brain& fee!bac$ normall allo"s for t"o
correcti#e mechanisms. Mne is the XonlineQ control an! correction of errors !uring the e+ecution
of a mo#ement. %he other is learning- the gra!ual a!aptation of motor comman!s& "hich ta$es
place after the e+ecution of one or more mo#ements.
In the B'Is base! on the operant con!itioning approach& fee!bac$ training is essential for the
user to ac8uire the control of his or her EEB response. %he B'Is base! on the pattern
recognition approach an! using mental tas$s !o not !efinitel re8uire fee!bac$ training.
>o"e#er& fee!bac$ can spee! up the learning process an! impro#e performance. Cursor control
has been the most popular tpe of fee!bac$ in B'Is. Gee!bac$ can ha#e man !ifferent effects&
some of them beneficial an! some harmful. Gee!bac$ use! in B'Is has similarities "ith
biofee!bac$& especiall EEB biofee!bac$.
20
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
BCI APPROACHES
Neuropsychological signals used in BCI applications.
Interfaces base! on brain signals re8uire on-line !etection of mental states fromspontaneous
acti#it- !ifferent cortical areas are acti#ate! "hile thin$ing !ifferent things;i.e. a mathematical
computation& an imagine! arm mo#ement& a music composition&etc...<. %he information of these
4mental states4 can be recor!e! "ith !ifferent metho!s..europschological signals can be
generate! b one or more of the follo"ingthree- implante! metho!s& e#o$e! potentials ;also
$no"n as e#ent relate! potentials<& an!operant con!itioning. Both e#o$e! potential an! operant
con!itioning metho!s arenormall e+ternall-base! BCIs as the electro!es are locate! on the
scalp. %able =.1!escribes the !ifferent signals in common use. It ma be note! that some of the
!escribe!signals fit into multiple categories. As an e+ample& single neural recor!ings ma
useoperant con!itioning in or!er to train neurons for control or ma accept the naturaloccurring
signals for control. *here this occurs& the signal is !escribe! un!er thecategor that best
!istinguishes it.
21
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
22
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
/=00-Interface
ERS/ERD-Interface
%he e#ent relate! snchroni)ation ( !esnchroni)ation is a signal tpe that can be measure!
"henthe sub:ect imagines a han! or foot mo#ement. ERS ( ERD "as researche! b a team at
thetechnical uni#ersit of Bra) in Austria NCAOEEBP. %he also !e#elope! an interface "hich
allo"s o#ement of a cursor in a t"o !imensional space b combining han! an! foot mo#ement.
After "o training sessions three of the fi#e test persons achie#e! a success rate bet"een J? an!
100percent& the t"o other persons onl L1an! E0 percent.%his coul! ha#e the cause in !ifferent
imaginations of han! mo#ement each person has.After a E2 training sessions "ith 1E0 trials a
2L-ear ol! paraplegic patient coul! mo#e the cursorpracticall error-free. Regar!ing his
immense han!icap the accomplishment of 0.?L letters perminute is !efinitel
2=
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
respectable.Another e+periment "as !one at the technical uni#ersit Bra). %he combine! the
ERS(ERDInterface"ith a GES ;functional electrical stimulation< "hich is the stimulation of
muscles using
surface electro!es. A patient "ho ha! lost his abilit to grasp "ith his han! is no" no longer
Ipe!e!.
2D
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
BCI RESEARCH: EXISTING SYSTEMS.
Different research groups "or$ on communication channels bet"een the brain an!
the computer. %he lea!ing groups are presente! in alphabetical or!er in %able =. 2. %hese
e+perimental interfaces inclu!e the har!"are use! in the BCI& the un!erling BCI
bac$en! soft"are& an! the user application. In assessing current sstems& se#eral factors
must be consi!ere!& inclu!ing fi#e mentione! b Ben Schnei!erman N=.1?P-
1. *hat is the time to learn the sstemY
2. *hat is the spee! of performanceY
=. >o" man an! "hat $in!s of errors !o users ma$eY
D. >o" "ell !o users maintain their $no"le!ge after an hour& a !a& or a "ee$Y
*hat is their retentionY
Berlin Brain-Computer-Interface (BBCI)
%he Berlin Brain-Computer-Interface is a :oint #enture of se#eral german research organisations.
'embers are-
K %he Institute Computer Architecture an! Soft"are %echnolog of the Graunhofer Societ
K %he research group Intelligent Data Analsis ;IDA<
K %he .europhsics Research Broup
K %he !epartment of .eurolog at the Campus Ben:amin Gran$lin of the CharitZ 0ni#ersit
'e!icine
K %he %echnical 0ni#ersit Berlin
%he BBCI pro:ect is sponsore! b the 'inistr for E!ucation an! Research of the Berman
go#ernment.
%he goal of the pro:ect is the !e#elopment of an EEB base! BCI sstem. %he applications of this
sstem are on the one han! computer supporte! "or$places& to control a cursor #ia brain "a#es
an!
on the other han! tools for paral)e! or paraplegic people.
%he BBCI pro:ect aims to shift the main learning effort to the computer. %herefore robust
artificial
learning an! signal processing algorithms nee! to be !e#elope! to classif an! interpret the brain
"a#es correctl. NBBCIP
BCI Classification Competitions
%he first Brain-Computer-Interface Competition too$ place at the Aaborator for Intelligent
Imaging an! .eural Computing of the Columbia 0ni#ersit in 2002.
2L
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
It "as initali)e! to [forster !e#elopment of machine learning techni8ues an! e#aluate !ifferent
algorithms for BCI6.
%he competition focuse! on classification an! signal processing algorithms. Data sets for se#eral
!ifferent BCI tas$ "ere pro#i!e! to be anal)e! b the participants.
%he rules "ere prett simple-
1. All !ata sets ma onl be e#aluate! single-trial
2. %he statistics(metrics outline! in the !escription of each !ataset shoul! be reporte!
=. B using the !ata sets the participant agrees to refer their origin
D. Cheating is uncool3
NBCICCO2002P
%he competition "as a great success. %herefore it "as repeate! at the 0ni#ersit of Bra)& Austria
in
200= an! at the %he Institute Computer Architecture an! Soft"are %echnolog of the Graunhofer
Societ in 200L. NBBCIP
Brain
1) The Brain Response Interface (Smith-Kettlewell Institute of Visual
Sciences in San Francisco).
SutterIs Brain Response Interface ;BRI< N=.JP uses #isuall e#o$e! potentials ;1E/Is<pro!uce! in
response to brief #isual stimuli. %hese E/Is are then use! to gi#e a !iscretecomman! to pic$ a
certain part of a computer screen. *or! processing output approaches10-12 "or!s(min. an!
accurac approaches ?0V "ith the use of epi!ural electro!es. %his the onl sstem mentione!
that uses implante! electro!es to obtain a stronger& less
contaminate! signal.
A BRI user "atches a computer screen "ith a gri! of ED smbols ;some of "hichlea! to other
pages of smbols< an! concentrates a gi#en smbol. A specific subgroup ofthese smbols
un!ergoes a e8uiluminant re!(green fine chec$ or plain color patternalteration in a simultaneous
stimulator scheme at the monitor #ertical refresh rate ;D0-@0frames(s<. Sutter consi!ere! the
usabilit of the sstem o#er time an! since coloralteration bet"een re! an! green "as almost as
effecti#e as ha#ing the monitor flic$er& hechose to use the color alteration because it "as sho"n
to be much less fatiguing for users.%his sstem is basicall the EEB #ersion of an ee mo#ement
recognition ssteman! contains similar problems because it assumes that the sub:ect is al"as
loo$ing at acomman! on the computer screen.
2) P3 Character Recognition (University of Illinois, USA).
2E
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
In a relate! approach& Gar"ell an! Donchin use the /= e#o$e! potential N=.1LP. A E+Egri!
containing letters from the alphabet is !isplae! on the computer monitor an! usersare as$e! to
select the letters in a "or! b counting the number of times that a ro" orcolumn containing the
letter flashes. Glashes occur at about 10 >) an! the !esire! letterflashes t"ice in e#er set of
t"el#e flashes. %he a#erage response to each ro" an!column is compute! an! the /= amplitu!e
is measure!. Response amplitu!e is reliabllarger for the ro" an! column containing the !esire!
letter. After t"o training sessions&users are able to communicate at a rate of 2.= characters(min&
"ith accurac rates of ?LV.%his sstem is currentl onl use! in a research setting.
3) ERS/ERD Cursor Control (University of Technology Graz, Austria)
/furtscheller an! his colleagues ta$e a !ifferent approach 0sing multiple electro!es place! o#er
sensorimotor corte+ the monitor
e#ent-relate! snchroni)ation( !esnchroni)ation ;ERS(ERD< NEDP. In all sessions&epochs "ith
ee an! muscle artifact are automaticall re:ecte!. %his re:ection can slo"!o"n sub:ect
performance. As this is a research sstem& the user application is a simplescreen that allo"s
control of a cursor in either the left or right !irection. In anothere+periment& for a single trial the
screen first appears blan$& then a target bo+ is sho"n onone si!e of the screen. A cross hair
appears to let the user $no" that he(she must begintring to mo#e the cursor to"ar!s the bo+.
Gee!bac$ ma be Delae! or imme!iate an!!ifferent e+periments ha#e slightl !ifferent
!isplas an! protocols. After t"o trainingsessions& three out of fi#e stu!ent sub:ects "ere able to
mo#e a cursor right or left "ithaccurac rates from J?-100V. 0nfortunatel& the other t"o
stu!ents performe! at E0Van! L1V. *hen a thir! categor "as a!!e! for classification&
performance !roppe! to a
lo" of E0V in the best case N=.2=P.
4) A Steady State Visual Evoked Potential BCI (Wright-Patterson Air Force
Base, The Air Force Research Laboratory, USA).
'i!!en!orf an! colleagues use operant con!itioning metho!s in or!er to train#olunteers to
control the amplitu!e of the stea!-state #isual e#o$e! potential ;SS1E/< toflorescent tubes
flashing at 1=.2L >) N=.11PN=.10PN=.?P. %his metho! of control ma beconsi!ere! as continuous
as the amplitu!e ma change in a continuous fashion. Either ahori)ontal light bar or au!io
fee!bac$ is pro#i!e! "hen electro!es locate! o#er theoccipital corte+ measure changes in signal
amplitu!e. If the 1E/ amplitu!e is belo" orabo#e a specifie! threshol! for a specific time
perio!& !iscrete control outputs aregenerate!. After aroun! E hours of training& users ma ha#e
2@
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
an accurac rate of greaterthan J0V in comman!ing a flight simulator to roll left or
right.Recogni)ing that the SS1E/ ma also be use! as a natural response&
5) Mu Wave Cursor Control (Wadsworth Center, Albany, USA).
*olpa" an! his colleagues free their sub:ects from being tie! to a flashing florescenttube b
training sub:ects to mo!if their mu "a#e N=.2DPN=.1P. %his metho! of control iscontinuous as the
mu "a#e ma be altere! in a continuous manner. It can be attenuate!b mo#ement an! tactile
stimulation as "ell as b imagine! mo#ement. A sub:ectIs maintas$ is to mo#e a cursor up or
!o"n on a computer screen. *hile not all sub:ects are ableto learn this tpe of biofee!bac$
control& the sub:ects that !o& perform "ith accuracgreater than or e8ual to ?0V. %hese
e+periments ha#e also been e+ten!e! to t"o!imensionalcursor mo#ement& but the accurac of
this is reporte! as ha#ing 6not reache!this le#el of accurac7 "hen compare! to the one-
!imensional control N=.11P.
.
6 An Implanted BCI (Georgia State University, USA).
%he implante! brain-computer interface sstem !e#ise! b ,enne! an!colleagues has been
implante! into t"o patients N=.2LPN=.@P. %hese patients are traine! to
control a cursor "ith their implant an! the #elocit of the cursor is !etermine! b the rateof
neural firing. %he neural "a#eshapes are con#erte! to pulses an! three pulses are aninput to the
computer mouse. %he first an! secon! pulses control T an! 9 position of thecursor an! a thir!
pulse as a mouse clic$ or enter signal.
%he patients are traine! using soft"are that contains a ro" of icons representingcommon phrases
;%al$ Assist !e#elope! at Beorgia %ech<. %here are t"o para!igmsusing this soft"are program
an! a thir! one using the #isual $eboar!. In the firstpara!igm& the cursor mo#es across the
screen using one group of neural signals an! !o"nthe screen using another group of larger
amplitu!e signals. Starting in the top left corner&the patient enters the leftmost icon. >e remains
o#er the icon for t"o secon!s so that thespeech snthesi)er is acti#ate! an! phrases are
pro!uce!. In the secon! para!igm& thepatient is e+pecte! to mo#e the cursor across the screen
from one icon to the other. %hepatient is encourage! to be as accurate as possible& an! then to
spee! up the cursormo#ement "hile attempting to remain accurate. In the thir! para!igm& a
#isual $eboar!is sho"n an! the patient is encourage! to spell his name as accuratel an!
8uic$l aspossible an! then to spell anthing else he "ishes. 0nfortunatel& the
ma+imumcommunication rate "ith this BCI has been aroun! = characters per minute.
2J
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
= AAS ;Amotrophic Aateral Sclerosis< is a fatal neuromuscular !isease characteri)e!
bprogressi#e muscle "ea$ness resulting in paralsis.
8) The Flexible Brain Computer Interface (University of Rochester, USA).
Baliss an! colleagues N=.2EP ha#e performe! an en#ironmental controlapplication in a #irtual
apartment that enables a sub:ect to turn on(off a light& tele#ision
set& an! ra!io or sa >i(Be to a #irtual person. %his sstem uses the /= e#o$e! potentialin an
immersi#e an! !namic 1irtual Realit "orl!. %he main !ra"bac$ of /=-base!BCIIs is their
slo"ness. Single trial analsis ma spee! up recognition& but often at thecost of accurac.
A single trial accurac a#erage of JLV "as obtaine! in an en#ironment of #irtual!ri#ing.
Sub:ects "ere instructe! to !ri#e in a #irtual to"n an! stop at re! stop lights"hile ignoring both
green an! ello" lights. %he sub:ects use! a #irtual realit helmet&an! a go cart "ith bra$e&
accelerator& an! steering output to control the #irtual car. *hilethis choice coul! ha#e cause!
more artifacts in the signal collection ;!ue to turning thesteering "heel an! bra$ing<& most of the
artifact !isco#ere! an! preprocesse! "as !ue toee mo#ement
APPLICATIONS
Introduction
Apart from being a non-con#entional input !e#ice for a computer "e ha#e foun! three main
application fiel!s for BCIs an! BCI relate! !e#ices "hich are more or less contro#ersial-
K 'e!ical applications
K >uman enhancement
K >uman manipulation
Medical applications
BCIs pro#i!e a ne" an! possibl onl communication channel for people suffering from se#ere
phsical !isabilities but ha#ing intact cogniti#e functions. Gor e+ample these !e#ices coul! help
in
treating ;or rather o#ercoming< paraplegia or amotrophia.
Some"hat relate! to this topic is the fiel! of .europrosthetics "hich !eals "ith constructing an!
surgicall implanting !e#ices use! for replacing !amage! areas of the brain an! more generall
for
neural !amages of an $in!.
Gor e+ample& the most "i!esprea! neuroprosthetic ;appro+. JL&000 people "orl!"i!e 200L
2?
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
N*I,IO./P< is the cochlear implant or bionic ear. %his !e#ice can help people "ith impaire!
hearing. In contrast to con#entional hearing ai!s this !e#ice is not a soun! amplifier but !irectl
stimulates an appropriate functioning ner#es.
Human enhancement
Definition
6>uman enhancement !escribes an attempt ;"hether temporar or permanent< to o#ercome the
current limitations of human cogniti#e an! phsical abilities& "hether through natural or artificial
means.7 N*I,IO>EP
>a#ing this !efinition in min! one can thin$ of man applications of the BCI in this fiel!. Gor
e+ample BCIs coul! help facilitate communication sstems in Cbernetic Mrganisms& Brain"a#e
Snchroni)ation& or e#en speculati#e things such as the E+ocorte+& among others.
Brain Computer Interfaces - 1= - Behm ( ,ollot)e$ ( >\s$e
Cbernetic Mrganism !escribes the enhancement of an organism b means of technolog. Gor
e+ample a BCI coul! enable the attachment of robotic limbs "ithout the use of the organismHs
original ner#ous sstem ;as long as the brain is intact<.
Brainwave Synchronization
6Brain"a#e snchroni)ation is the practice to entrain oneIs brain"a#es to a !esire! fre8uenc& b
means of a perio!ic stimulus "ith correspon!ing fre8uenc. %he stimulus can be aural as in the
case
of binaural beats& or #isual& as "ith a Dreamachine.7 N*I,IOBSP
A BCI coul! be use! to !etect an! classif the current state of min! of an in!i#i!ual an! acti#el
a!:ust the fre8uenc of the Brain"a#e Snchroni)ation to achie#e a certain state of min!.
Excortex
6An e+ocorte+ ;speculati#e< is an e+ternal information processing sstem that augments& in a
subtle
an! seamless fashion #ia a brain-computer interface& the brainIs biological high-le#el cogniti#e
processes.7 N*I,IOECP
Human manipulation
%he notion that a BCI coul! allo" a t"o-"a communication bet"een a human an! a computer
gi#es rise to more contro#ersial potential uses of such a !e#ice. 0sing such a communication
mechanism one coul! imagine !irectl influencing an in!i#i!ualHs thoughts& !ecisions& emotions
or
thin$ing. Mf course& the mere 6rea!ing7 of the min! coul! be put to criminal use& e.g. un"ante!
=0
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
rea!ing of pass"or!s& locations& etc.
*hile this ma soun! li$e science-fiction& metho!s researche! in social-pscholog such as
a!#ertisement& me!ia manipulation& propagan!a& group !namics or peer pressure ha#e been
pro#en
to be successful in altering an in!i#i!ualHs beha#ior an! their effecti#eness is un!ispute!. So& in
principle manipulation is possible.
E+periments to e+plore possibilities of min!-control are b no means theoretical an! inclu!e the
use
of !rugs an! electronic signals to alter brain functioning. Gor e+ample the pro:ect co!ename!
',0A%RA con!ucte! b the CIA !ates bac$ to the 1?L0Hs an! "as aime! at researching
min!control
N*I,IO',P.
Brain-computer-interfaces present a ne" le#el of technolog that coul! be use! to acti#el
manipulate an in!i#i!ual.
%he B'I technologies of to!a can be bro$en into three ma:or areas-
1. Au!itor an! #isual prosthesis
- Cochlear implants
- Brainstem implants
- Snthetic #ision
- Artificial silicon retina
2. Gunctional-neuromuscular stimulation ;G.S<
G.S sstems are in e+perimental use in cases "here spinal cor! !amage or a stro$e has se#ere!
the lin$ bet"een brain an! the peripheral ner#ous sstem. %he can use brain to control their
o"n limbs b this sstem
=. /rosthetic limb control
%hought controlle! motori)e! "heel chair.
=1
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
%hought controlle! prosthetic arm for amputee.
1arious neuroprosthetic !e#ices
Mther #arious applications are 'ental 'ouse Applications in technolog pro!ucts& e.g.& a mobile
phone attachment that allo"s a phsicall challenge! user to !ial a phone number "ithout
touching it or spea$ing into it. Sstem lets ou spea$ "ithout saing a "or! in effecti#e 1E
]construction of unmanne! sstems& in space missions& !efense areas etc. .ASA an! DAR/A
ha#e use! this technolog effecti#el. Communication o#er internet can be mo!ifie!.
The Mental Typewriter:
'arch 1D& 200E Scientists !emonstrate! a brain-computer interface that translates brain signals
into computer control signals this "ee$ at CeBI% in Berlin. %he initial pro:ect !emonstrates ho"
a paralse! patient coul! communicate b using a mental tpe"riter alone K "ithout touching
the $eboar!. In the case of serious acci!ent or illness& a patientHs limbs can be paral)e!&
se#erel restricting communication "ith the outsi!e "orl!. %he interface is alrea! sho"ing ho"
it can help these patients to "rite te+ts an! thus communicate "ith their en#ironment It "ill be
some ears& though& before the mental tpe"riter can be use! in e#er!a applications. Gurther
research is nee!e!& in particular to refine the EEB sensors.
BCI offers paralyzed patients improved quality of life:
%uebingen& Berman. A brainKcomputer interface installe! earl enough in patients "ith neuron-
!estroing !iseases can enable them to be taught to communicate through an electronic !e#ice
an! slo" !estruction of the ner#ous sstem %he research focuses on a con!ition calle! the
completel loc$e!-in state ;CAIS& a total lac$ of muscle control<. In a CAIS situation& intentional
thoughts an! imager can rarel be acte! upon phsicall an!& therefore& are rarel follo"e! b a
stimulus. %he research suggests that as the !isease progresses an! the probabilit for an e+ternal
e#ent to function as a lin$ bet"een response an! conse8uence becomes progressi#el smaller it
ma e#entuall #anish altogether.
Researchers ha#e foun! that b implementing a brain-computer Kinterface before the completel
loc$e!-in state occurs& a patient can be taught to communicate through an electronic !e#ice "ith
great regularit. %he continue! interaction bet"een thought& response an! conse8uence is
belie#e! to slo" the !estruction of the ner#ous sstem.
%he fin!ings are also raising a number of ne" 8uestions about the 8ualit of life amongst
paralsis sufferers. /atients sur#ee! "ere foun! to be much healthier mentall than
pschiatricall !epresse! patients "ithout an life-threatening bo!il !isease. Mnl ?V of AAS
=2
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
patients sho"e! long episo!es of !epression an! most "ere !uring the perio! follo"ing
!iagnosis an! a perio! of "ee$s after tracheotom
Adaptive BCI for Augmented Cognition and
Action :
%he goal of this pro:ect is to !emonstrate impro#e! human(computer performance for specific
tas$s through !etection of tas$-rele#ant cogniti#e e#ents "ith real-time EEB ;Gig. 1<. Gor
e+ample& in tas$s for "hich there is a !irect tra!eoff bet"een reaction time an! error rate& ;such
as tping or #isual search< it ma be beneficial to correct a userHs errors "ithout interrupting the
pace of the primar tas$. Such a user interface is possible through the !irect !etection of EEB
signatures associate! "ith the perception of a error& often referre! to as Error Relate! .egati#it.
In general such signatures ma be use! to !namicall a!:ust the beha#ior of human-computer
interfaces an! information !isplas.
%his pro:ect a!#ances signal analsis techni8ues for high !ensit EEB to !etect !iscrete e#ents
associate! "ith cogniti#e processing. Correspon!ing real-time a!apti#e interfaces "ith sub-
secon! latenc are being !esigne! to e#aluate this concept of an a!apti#e brain-computer
interface in three specific applications (1) Error and conflict perception:
Error relate! negati#it ;ER.< in EEB has been lin$e! to percei#e! response errors an! conflicts
in !ecision-ma$ing. In this pro:ect "e ha#e !e#elope! single trial ER. !etection to pre!ict tas$-
relate! errors. %he sstem can be use! as an automate! real-time !ecision chec$er for time-
sensiti#e control tas$s. In the first phase of this pro:ect "e !emonstrate! impro#e!
human(computer performance at a rapi! force! choice !iscrimination tas$ "ith an a#erage 2=V
re!uction of human errors ;results on one sub:ect are sho"n in Gig. 2<. %his open-loop error
correction para!igm represents the first application of real-time cogniti#e e#ent !etection an!
==
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
!emonstrates the utilit of real-time EEB brain monitoring "ithin the Augmente! Cognition
program. *e "ill e#aluate #i!eo game scenarios "ith close!-loop fee!bac$ at latencies of less
than 1L0 ms "here !etecte! errors are correcte! or application parameters such as spee! are
#arie! accor!ing to the measure! or 4gauge!4 conflict perception.
(2) Working memory encoding.
%ransient mo!ulation of oscillations in the theta ;D-J >)< an! gamma ;20-=0 >)< ban!s&
recor!e! using EEB an! magnetoencephalograph ;'EB<& ha#e been implicate! in the enco!ing
an! retrie#al of semantic information in "or$ing memor. In this pro:ect "e "ill e+ploit these
neural correlates of semantic processing to !etect problems "ith semantic information
processing. %his memor gauge coul! be use! to !etect memor recall !eficits& an! repeat or
enhance the presente! information an! thus better prime memor recall.
(3) Rapid visual recognition:
*e are e+ploring the signals elicite! b #isual target !etection& "hich "ere recentl obser#e! in
rapi! se8uential #isual presentation ;RS1/< e+periments. *e ha#e !emonstrate! that the
!etection of these signals on a single trial basis can be use! to replace the slo" manual response
of a human operator& thereb significantl increasing the throughput of image search tas$s ;Gig
=<. %his para!igm has the potential to impro#e the performance of Image Analsts "ho nee! to
routinel sur#e large #olumes of aerial imager "ithin short perio!s of time. In a!!ition& the
approach loo$s to measure the 4bottlenec$4 bet"een constant !ela perceptual processing an!
more #ariable !ela cogniti#e processing. %hus the !etecte! signatures can be use! to 4gauge4 if
cogniti#e sstems are capable(incapable of assimilating perceptual input for fast !ecision
ma$ing.
In the first phase of this pro:ect a full automate! real-time signal analsis sstem an! har!"are
infrastructure has been !e#elope! that can gi#e short latenc fee!bac$ to the user "ithin L0ms of
the recor!e! acti#it. %he signal processing sstem a!apti#el learns to !etect e#o$e! responses
from the real-time streaming EEB signal. %he current sstem& "hich is use! for tas$s 1 an! =&
can be
configure! for single trial !etection for an number of cogniti#e e#ents such ER.& rapi! #isuual
recognition& rea!iness potential& response to o!!ball stimulus ;/=00<& as "ell as con#entional
#isual& au!itor& or somato-sensor responses. *e are in the progress of appling this sstem to
e#ent !etection in the *arship Comman!er - a common tas$ set propose! for integration an!
e#aluation b the Augmente! Cognition /rogram
=D
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
=L
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
E+perimental Brain Computer Interface Soft"are for the 'o!ular EEB ;%he ABI soft"are<
E.1. Intro!uction-
ABI is a simple soft"are for the 'o!ular EEB that implements an e+perimental Brain Computer
Interface ;BCI<. .o"a!as& BCI research is an highl acti#e fiel!& but the e+isting
technolog is still immature for its use outsi!e of a labIs settings. %he ABI soft"are tries
to pro#i!e a simple tool for hobbists to !o e+periments on its o"n "ith BCI technolog.
Work of the software:
%he ABI is a BCI base! on trials. A trial is a time inter#al "here the user generates brain"a#es
to perform an action. %he BCI tries to process this signal an! to associate it to a gi#en class. %he
association is !one b fee!ing a neural net "ith the preprocesse! EEB !ata. %he neural netIs
output is then further processe! an! this final output correspon!s to the gi#en class. %he neural
net shoul! be traine! in or!er to learn the association.
%he classifierIs i!ea is hea#il base! on Christin SchWferIs !esign ;"inner of the BCI
Competition II& 'otor Imaginer %rials<.
%he ABI soft"are allo"s ou to
1 Do simple Biofee!bac$. 9ou can !ispla ra" EEB channels& narro" ban! fre8uenc
amplitu!es an! classes.
2 Simulate trials.
= Recor! trials for a number of choice of !ifferent classes.
D %rain the interface.
6.3. The classification achieved by this software:
%he metho! has been pre#iousl applie! to the !ata pro#i!e! b the BCI Competition II !ata
;!ataset III& Bra) 0ni#ersit& 'otor Imaginar< an! compare! against the results obtaine! b the
contributors. %he metho! has outperformed the results achie#e! b them& obtaining a higher
'utual Information ;"hich "as the criterion use! in the competition< of 0.E@ bits ;the "inner of
the competition obtaine! 0.E1 bits<.
Mf course& it is #er important that more people test the soft"are an! report its results to impro#e
the metho!. Statistical stabilit can onl be guarantee! if more people tr it out.
6.4. Instructions:
=E
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
B e+ecuting ABI& it rea!s a configuration file calle! 4abi.t+t4 ;"hich ou can e!it "ith a simple
te+t e!itor<& "here the "a the BCI shoul! act is specifie!. ABI tries to loa! the trial file !efine!
in the configuration file. %he trial file is a te+t !atabase containing trials for !ifferent classes.
ABI has three operating mo!es- SIMULATION& RECORDING an! TRAINING. 9ou can
s"itch bet"een operating mo!es b pressing F1& F2 or F3 respecti#el ;the soft"are !oesnIt
change its mo!e instantl& because a trial shoul!nIt be interrupte! in the mi!!le<.
%he operation is 8uite simple. %he user recor!s se#eral trials for the !ifferent classes
;RECORDING mo!e<. Each class is associate! to a !ifferent mental tas$. After recor!ing a
reasonable amount of trials ;more than L0 trials for each class<& the user can train the sstem to
learn a "a to !iscriminate bet"een the !ifferent classes ;TRAINING mo!e<. %his process can
be repeate! in or!er to impro#e the 8ualit of the recognition. %he sstem can be teste! un!er the
SIMULATION mo!e.
An e+planation of the !ifferent mo!es follo"s.
E.D.1.SI'0AA%IM. an! RECMRDI.B
%hese t"o mo!es perform single trials. %he SIMULATION mo!e is use! to test the BCI.
RECORDING is the same as SI'0AA%IM.& "ith the !ifference that the EEB !ata is recor!e!
an! use! as training e+amples. A trial has the follo"ing structure
=@
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
As ou can see& a trial is compose! of three subinter#als& "hose !uration is !efine! b the
#ariables TPreparation& TPrerecording an! TrialLength& in the configuration file.
. %RAI.I.B
/ressing the F3 $e& the sstem starts to train the neural net "ith the a#ailable !ata. %he training set use! for this purpose is the
set of the last TrialBuffer recor!e! trialsI features. E+ample- Suppose ou ha#e recor!e! =00 trials& an! TrialBuffer 100
%hen the sstem e+tracts the features of the 100 last recor!e! trials to form the training set.
%raining time !epen!s upon the comple+it of the training !ata an! the amount of recor!e! !ata. %he training !ata is not al"as
separable. If the mental tas$ for class 1 is too similar to the mental tas$ for class 2& then the neural net "onIt be able to !o the
separation- this isnIt magic --< .
considerations
BCI technolog is still in its infanc& so little is $no"n about "hich mental tas$s are better than
others for BCIs. Also& the electro!e placing is important. If our electro!e setting isnIt
=J
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
appropiate& then it can happen that the e#en arenIt recor!ing the cortical areas relate! to the
mental tas$3
Research has !isco#ere! the follo"ing changes in electrical acti#it !uring mental tas$s ;this list
isnIt complete& I hope that the MpenEEB communit "ill !isco#er some more<-
1 Motor Imaginery: Imagination of phsical mo#ement pro!uces changes in the
sensormotor corte+. In e+ample& imagination of left an! right mi!!le finger imagination
pro!uces changes& namel ;!e-<snchroni)ation on electro!e positions aroun! C= an! CD. Boo!
features are aroun! 10 an! 20 >).
2 Rotation of 3D objects: Aiterature state! that !uring imagination of rotation of =!
ob:ects in#ol#es frontal an! temporal lobe acti#it. %he seem to sinchroni)e. Boo! features are
aroun! 10 >).
=?
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
CONCLUSION
Present and Future.
%he practical use of BCI technolog !epen!s on an inter!isciplinar cooperationbet"een
neuroscientists& engineers& computer programmers& pschologists& an!
rehabilitation specialists& in or!er to !e#elop appropriate applications& to i!entifappropriate
users groups& an! to pa careful attention to the nee!s an! !esires of
in!i#i!ual users.%he prospects for controlling computers through neural signals are
in!ee!!ifficult to :u!ge because the fiel! of research is still in its infanc. 'uch progress
hasbeen ma!e in ta$ing a!#antage of the po"er of personal computers to perform theoperations
nee!e! to recogni)e patterns in biological impulses& but the search for ne" an!more useful
signals still continues.If the a!#ances of the 21st centur match the stri!es of the past fe"
!eca!es& !irectneural communication bet"een humans an! computers ma ultimatel mature
an! fin!"i!esprea! use. /erhaps ne"l purchase! computers "ill one !a arri#e "ith
biologicalsignal ensors an! thought-recognition soft"are built in& :ust as $eboar! an! mouse
arecommonl foun! on to!aIs units. Cultures ma ha#e !i#erse ethics& but regar!less& in!i#i!ual
liberties an! human life are al"as #alue! o#er an! abo#e machines. *hat happens "hen
humans merge "ith machinesY %he 8uestion is not "hat "ill the computer be li$e in the future&
but instea!& "hat "ill "e be li$eY *hat $in! of people are "e becomingY
B'IQs "ill ha#e the abilit to gi#e people bac$ their #ision an! hearing. %he "ill also change
the "a a person loo$s at the "orl!. Some!a these !e#ices might be more common than
$eboar!s. Is someone "ith a snthetic ee& less a person than someone "ithoutY Shall "e
process signals li$e ultra#iolet& T-ras& or ultrasoun!s as robots !oY %hese 8uestions "ill not be
ans"ere! in the near future& but at some time the "ill ha#e to be ans"ere!. *hat an interesting
!a that "ill be. An! further still-
In principle "e coul! !o this for all the senses5recor! not :ust "hatou see& but also "hat ou
hear& taste& smell& an! feel& all at the le#elof our brain. /laing bac$ such an e+perience "oul!
be a little li$ereli#ing it. 9ou might e#en be able to pla that $in! of sensor recor!ingbac$ for
someone else& turning the e+perience ou ha! into a set ofner#e impulses that coul! be sent into
the other personHs brain& allo"inghim or her to e+perience at least the sensor parts of an e#ent
from
our perspecti#e. . . .
D0
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
ANNEXURE 1: BIOMEDICAL CONSIDERATIONS.
%he brain contains appro+imatel 100 billion neurons NA1-1P. Each neuron ma ha#e
as man as 2000 or more connections "ith other neurons an! ma recei#e as man as
20&000 inputs. Abstractl& a neuron is probabl the most !i#erse& in terms of form an!
si)e& of all cells in the bo!2 ho"e#er& all neurons ha#e in common the functional
properties of integration& con!uction an! transmission of ner#e impulses. A neuron
consists of three basic parts ;Gig. A1-1<-
U Cell body ;or soma<. %his main part has all of the necessar components of the
cell& such as the nucleus ;contains D.A<& en!oplasmic reticulum an! ribosomes
;for buil!ing proteins< an! mitochon!ria ;for ma$ing energ<. If the cell bo!
!ies& the neuron !ies.
U Axon - %his long& cable-li$e pro:ection of the cell carries the electrochemical
message ;ner#e impulse or action potential< along the length of the cell.
Depen!ing upon the tpe of neuron& a+ons can be co#ere! "ith a thin laer of
melin& li$e an insulate! electrical "ire. 'elin is ma!e of fat& an! it helps to
spee! transmission of a ner#e impulse !o"n a long a+on. 'elinate! neurons are
tpicall foun! in the peripheral ner#es ;sensor an! motor neurons<& "hile nonmelinate!
neurons are foun! in the brain an! spinal cor!.
U Dendrites or nerve endings - %hese small& branch-li$e pro:ections of the cell
ma$e connections to other cells an! allo" the neuron to tal$ "ith other cells or
percei#e the en#ironment. Den!rites can be locate! on one or both en!s of the
cell. %here are t"o tpes of !en!rites& apical an! basal !en!rites.
D1
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
Figure A1-1. .euron topolog.
Each neuron is in contact through its a+on an! !entrites "ith other neurons& so that each neuron
is an
interconnecting segment in the net"or$ of the ner#ous sstem.
L@
.eurons come in man si)es. Gor e+ample& a single sensor neuron from the
fingertip has an a+on that e+ten!s the length of the arm& "hile neurons "ithin the brain
ma e+ten! onl a fe" millimeters. .eurons ha#e !ifferent shapes !epen!ing on "hat
the !o. 'otor neurons that control muscle contractions ha#e a cell bo! on one en!& a
long a+on in the mi!!le an! !en!rites on the other en!2 sensor neurons ha#e !en!rites
on both en!s& connecte! b a long a+on "ith a cell bo! in the mi!!le. .eurons also #ar
"ith respect to their functions-
U Sensory neurons carr signals from the outer parts of the bo! ;peripher< into
the central ner#ous sstem.
U Motor neurons ;motoneurons< carr signals from the central ner#ous sstem to
the outer parts ;muscles& s$in& glan!s< of the bo!.
U Receptors sense the en#ironment ;chemicals& light& soun!& touch< an! enco!e this
information into electrochemical messages that are transmitte! b sensor
neurons.
U Interneurons connect #arious neurons "ithin the brain an! spinal cor!.
Figure A1-2. Some tpes of neurons- interneuron& sensor neuron& motoneuron an! cortical
prami!al cell.
%he snapse& a speciali)e! site of contact bet"een neurons& is of prime
D2
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
significance in the integrati#e acti#ities of the ner#ous sstem& as the information from
one neuron flo"s to another neuron across of it. %he snapse is a small gap separating 2
neurons& an! consists of-
LJ
1. A presnaptic en!ing that contains neurotransmitters& mitochon!ria an! other cell
organelles.
2. A postsnaptic en!ing that contains receptor sites for neurotransmitters.
=. %he snaptic cleft- a space bet"een the presnaptic an! postsnaptic en!ings.
Gor communication bet"een neurons& an electrical impulse must tra#el !o"n an a+on
to the snaptic terminal.
Electric potentials are pro!uce! at the snaptic :unctions& "hich can be locali)e! o#er
the a+on ;a+oa+onic snapse<& the soma ;a+osomatic snapse< or the !entrites
;a+o!en!ritic snapse< ;Gig A1.=<& an! reflect the communication bet"een neurons.
*hen a neurotransmitter bin!s to a receptor on the postsnaptic si!e of the snapse& it
changes the postsnaptic cellIs e+citabilit- it ma$es the postsnaptic cell either more or
less li$el to fire an action potential. If the number of e+citator postsnaptic e#ents are
large enough& the "ill cause an action potential in the postsnaptic cell an! the
continuation of the 4message.4 In a!!ition to %he EEB pic$ up this snchroni)e!
subthreshol! !entritic potentials pro!uce! b the postsnaptic acti#it of man neurons
summe! NA1-2P.

Figure A1-3. %pes of snapses. Grom left t"o right- a+oa+onic snapse& a+o!en!ritic snapse&
an!
D=
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
a+osomatic snapse.
.ot all tpes of brain acti#it ha#e i!entical impact on the EEB. %he !epth&
orientation an! intrinsic smmetr of connections in the corte+ are significant in it. %he
primar cell of importance in the neocorte+ is the prami!al cell. ItIs $no"n that its
neurotransmitter is a potent e+citator transmitter. %he prami!al cell recei#es man
inputs from stellate cells that are also e+citator. %he prami!al cell is !ifferent from
other neurons in that it #iolates one of the fun!amental rules of stan!ar!
neurophsiolog& that onl a+ons pro!uce action potentials "hich transmit information
from one cell to another& an! !en!rites pro!uce e+citator an! inhibitor slo" potentials
that summate at the a+on hilloc$ "here the establish the action potentials. In the case of
the prami!al cell& the apical !en!rite& "hich is a long shaft bet"een the basal an! apical
region& can actuall pro!uce action potentials& an! these in turn act to amplif the
po"erful action potentials that no" pro:ect to output sstems--sensor& motor&
autonomic& or integrati#e NA1.=P. Gor this reason& prami!al cells are thought to cause the
strongest part of the EEB signal
THE HUMAN BRAIN
All of it happens in the brain. %he brain is un!oubte!l the most comple+ organ foun! among the
carbon-base! life forms. So comple+ it is that "e ha#e onl #ague information about ho" it
"or$s. %he a#erage human brain "eights aroun! 1D00 grams. %he most rele#ant part of brain
concerning B'IQs is the cerebral corte+. %he cerebral corte+ can be !i#i!e! into t"o
hemispheres. %he hemispheres are connecte! "ith each other #ia corpus callosum. Each
hemisphere can be !i#i!e! into four lobes. %he are calle! frontal& parietal& occipital an!
temporal lobes. Cerebral corte+ is responsible for man higher or!er functions li$e problem
sol#ing& language comprehension an! processing of comple+ #isual information. %he cerebral
corte+ can be !i#i!e! into se#eral areas& "hich are responsible of !ifferent functions. %his $in!
of $no"le!ge has been use! "hen "ith BCIQs base! on the pattern recognition approach. %he
mental tas$s are chosen in such a "a that the acti#ate !ifferent parts of the cerebral corte+.
Cortical Area Gunction
Au!itor Association Area /rocessing of au!itor information
Au!itor Corte+ Detection of soun! 8ualit ;lou!ness& tone<
Speech Center ;BrocaQs area< Speech pro!uction an! articulation
/refrontal Corte+ /roblem sol#ing& emotion& comple+ thought
'otor Association Corte+ Coor!ination of comple+ mo#ement
/rimar 'otor Corte+ Initiation of #oluntar mo#ement
/rimar Somatosensor Corte+ Recei#es tactile information from the bo!
DD
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
Sensor Association Area /rocessing of multisensor information
1isual Association Area Comple+ processing of #isual information
*ernic$eQs Area Aanguage comprehension
PRINCIPLES OF ELECTROENCEPHALOGRAPHY.
The Nature of the EEG signals.
%he electrical nature of the human ner#ous sstem has been recogni)e! for more
than a centur. It is "ell $no"n that the #ariation of the surface potential !istribution on
the scalp reflects functional acti#ities emerging from the un!erling brain N2.1P. %his
surface potential #ariation can be recor!e! b affi+ing an arra of electro!es to the scalp&
an! measuring the #oltage bet"een pairs of these electro!es& "hich are then filtere!&
amplifie!& an! recor!e!. %he resulting !ata is calle! the EEB. Gig. 1-1 sho"s "a#eforms
of a 10 secon! EEB segment containing si+ recor!ing channels& "hile the recor!ing sites
are illustrate! in Gig. 2-2. In our e+periments& "e ha#e use! the10-20 Sstem of
Electro!e /lacement& "hich is base! on the relationship bet"een the location of an
electro!e an! the un!erling area of cerebral corte+ ;the 4104 an! 4204 refer to the 10V
or 20V interelectro!e !istance< N2.@P.
Figure
DL
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
Figure 2-2. %he 10-20 Sstem of Electro!e /lacement.
Each site has a letter ;to i!entif the lobe< an! a number or another letter to i!entif the
hemisphere
location. %he letters G& %& C& /& an! M stan! for Grontal& %emporal& Central& /arietal an!
Mccipital. ;.ote
that there is no 4central lobe4& but this is :ust use! for i!entification purposes.< E#en numbers
;2&D&E&J< refer
to the right hemisphere an! o!! numbers ;1&=&L&@< refer to the left hemisphere. %he ) refers to an
electro!e
place! on the mi!line.
Nasion: point bet"een the forehea! an! nose.
Inion: Bump at bac$ of s$ull
%he EEB is thought to be the snchroni)e! subthreshol! !entritic potentialspro!uce! b the
snaptic acti#it of man neurons summe! N2.2P. In its formation not all
tpes of brain acti#it ha#e i!entical impact. %he !epth& orientation an! intrinsicsmmetr of
connections in the corte+ are significant in it. As it is e+pose! in pre#ious
"or$s N2.2PN2.=P& prami!al cells are thought to cause the strongest part of the
EEBsignal1..o"a!as& mo!ern techni8ues for EEB ac8uisition collect these
un!erlingelectrical patterns from the scalp& an! !igitali)e them for computer storage.
Electro!escon!uct #oltage potentials as micro#olt le#el signals& an! carr them into amplifiers
thatmagnif the signals appro+imatel ten thousan! times. %he use of this technolog!epen!s
strongl on the electro!es positioning an! the electro!es contact. Gor this reason&electro!es are
usuall constructe! from con!ucti#e materials& such us gol! or sil#erchlori!e& "ith an
DE
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
appro+imati#e !iameter of 1 cm& an! sub:ects must also use acon!ucti#e gel on the scalp to
maintain an acceptable signal to noise ratio.
EEG wave groups.
%he analsis of continuous EEB signals or brain "a#es is comple+& !ue to thelarge amount of
information recei#e! from e#er electro!e. As a science in itself& it has tobe complete! "ith its
o"n set of perple+ing nomenclature. Different "a#es& li$e so manra!io stations& are categori)e!
b the fre8uenc of their emanations an!& in some cases&b the shape of their "a#eforms.
Although none of these "a#es is e#er emitte! alone& thestate of consciousness of the in!i#i!uals
ma ma$e one fre8uenc range morepronounce! than others. Gi#e tpes are particularl
important-
BETA. %he rate of change lies bet"een 1= an! =0 >)& an! usuall has a lo" #oltage
bet"een L-=0 1 ;Gig. 2-E<. Beta is the brain "a#e usuall associate! "ith acti#e
thin$ing& acti#e attention& focus on the outsi!e "orl! or sol#ing concrete problems. It can
reach fre8uencies near L0 hert) !uring intense mental acti#it.
Figure 2-4. Alpha ;left< an! Beta ;right< "a#es.
ALPHA. %he rate of change lies bet"een J an! 1= >)& "ith =0-L0 1 amplitu!e ;Gig 2-
D<. Alpha "a#es ha#e been thought to in!icate both a rela+e! a"areness an! alsoinattention.
%he are strongest o#er the occipital ;bac$ of the hea!< corte+ an! also o#erfrontal corte+. Alpha
is the most prominent "a#e in the "hole realm of brain acti#it an!possibl co#ers a greater
range than has been pre#iousl thought of. It is fre8uent to see apea$ in the beta range as high as
20 >)& "hich has the characteristics of an alpha staterather than a beta& an! the setting in "hich
such a response appears also lea!s to the sameconclusion. Alpha alone seems to in!icate an
empt min! rather than a rela+e! one& amin!less state rather than a passi#e one& an! can be
D@
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
re!uce! or eliminate! b opening theees& b hearing unfamiliar soun!s& or b an+iet or mental
concentration.
THETA. %heta "a#es lie "ithin the range of D to @ >)& "ith an amplitu!e usuall greaterthan 20
U1. %heta arises from emotional stress& especiall frustration or !isappointment. %heta has been
also associate! "ith access to unconscious material& creati#e inspirationan! !eep me!itation. %he
large !ominant pea$ of the theta "a#es is aroun! @ >).
DELTA. Delta "a#es lie "ithin the range of 0.L to D >)& "ith #ariable amplitu!e. Delta"a#es
are primaril associate! "ith !eep sleep& an! in the "a$ing state& "ere thought toin!icate
phsical !efects in the brain. It is #er eas to confuse artifact signals cause! bthe large
muscles of the nec$ an! :a" "ith the genuine !elta responses. %his is becausethe muscles are
near the surface of the s$in an! pro!uce large signals "hereas the signal"hich is of interest
originates !eep in the brain an! is se#erel attenuate! in passingthrough the s$ull. .e#ertheless&
"ith an instant analsis EEB& it is eas to see "hen theresponse is cause! b e+cessi#e
mo#ement.
Figure 2-6. Delta "a#e.
GAMMA. Bamma "a#es lie "ithin the range of =L>) an! up. It is thought that this ban!reflects
the mechanism of consciousness - the bin!ing together of !istinct mo!ular brainfunctions into
coherent percepts capable of beha#ing in a re-entrant fashion ;fee!ing bac$on themsel#es o#er
time to create a sense of stream-of-consciousness<.
MU. It is an J-12 >) spontaneous EEB "a#e associate! "ith motor acti#ities an!ma+imall
recor!e! o#er motor corte+ ;Gig. 2-J<. %he !iminish "ith mo#ement or thentention to mo#e.
'u "a#e is in the same fre8uenc ban! as in the alpha "a#e ;Gig. 2-@<& but this last one is
recor!e! o#er occipital corte+.
DJ
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
Figure 2-7. 'u ;left< an! alpha ;right< "a#es.
'ost attempts to control a computer "ith continuous EEB measurements "or$ bmonitoring
alpha or mu "a#es& because people can learn to change the amplitu!e of theset"o "a#es b
ma$ing the appropriate mental effort. A person might accomplish thisresult& for instance& b
recalling some strongl stimulating image or b raising his or herle#el of attention.
D?
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
REGERE.CES-
E. R. ,an!el& F. >. Sch"art)& an! %. '. Fessell& /rinciples of .eural Science& =r!
e!. .e" 9or$- Else#ier(.orth->ollan!& 1??1
A!am ,eiper &%he Age of .euroelectronics
Ale+an!er Behm& 'ar$us A. ,ollot)e$ an! Gabian >\s$e. Controlling
computers b thoughts. http-((spectrum.ieee.org(biome!ical(!e#ices(the-
brainmachine-interface-unplugge!(0
B.*.Fer#is& E.C.Ifeachor& an! E.'.Allen. %he remo#al o ocular artifacts from the
BCI Competition& http-((liinc.bme.columbia.e!u(competition.htm
Berger http-((en."i$ipe!ia.org("i$i(>ansOBerger
Berlin Brain-Computer-Interface& http-((bbci.!e(
Brain-Computer Interface %echnolog- %heor an! /ractice-Girst International
'eeting /rogram an! /apers& %he Rensselaer#ille Institute& Rensselaer#ille& .e"
9or$
Brain-Computer Interface&
http-((en."i$ipe!ia.org("i$i(BrainOcomputerOinterface
Dr Aarr Abbott . Electrical properties of neurons
electroencephalogram -a re#ie" . http-((ne"ton.bme.columbia.e!u("olpa".html
Electroencephalograph& pages 1L-2E
E+ocorte+ & http-((en."i$ipe!ia.org("i$i(E+ocorte+
G. >. Aopes !a Sil#a an! A. #an Rotter!am. Biophsical aspects of EEB an!
'EB generation. In E. .ie!ermeer an! G. >. Aopes !a Sil#a& e!itors&
B. Bratton& '.B.>. Coles an! E. Donchin. A .e" 'etho! for off K line remo#al
of Mcular artifacts. Electroencephalograph an! .europhsiolog& 1ol LL& pp.
DEJ-DJD&1?J=
>ansBra) Brain-Computer-Interface& http-((bci.tugra).at(in!e+.html
L0
Dept. of ECE College of Engineering
Brain Computer Interfacing Seminar Report 2011
>e"ett& Baec$er& Car!& Care& Basen& 'antei& /erlman& Strong an!
Forge Ba)tarrica Mchoa&Bar Barcia 'olina.%oura!: Ebrahimi. EEB Signal
Classification for BrainComputer Interface Applications&
'i!!en!orf& '.S.& 'c'illan& B.& Calhoun& B.& un! Fones& ,.S.&1???&Brain-
computer interfaces base! on the stea!-state #isual-e#o$e! response
.euroimaging& http-((en."i$ipe!ia.org("i$i(.euroimaging
Schal$& Dennis F. 'cGarlan!& an! Fonathan R. *olpa". BRAI.-CM'/0%ER
I.%ERGACES GMR CM''0.ICA%IM. A.D CM.%RMA
1erplan$. >uman-Computer Interaction& http-((sigchi.org(c!g(c!g2.html
L1
Dept. of ECE College of Engineering

Anda mungkin juga menyukai