3. METODOLOGÍA DE LA INVESTIGACIÓN
3.5. Análisis situacional
3.5.3. Caracterización de elementos culinarios de la industria de alimentos y
The existene of eletrial ativity in the brain was disovered more than
a entury ago by Rihard Carton [64℄. However, it was not until the early
1920s when EEG was reorded from the human salp for the rst time [65℄.
Nowadays, EEG has beome the most prevalent method for reording brain
ativity for BCI systems.
SalpEEGreordingdisplaysthe diereneineletrialpotentialsbetween
twodierentsites onthehead,superimposingtheerebral ortexthatisloser
tothe reordingeletrode. Theproblemoftheextremelylowamplitudevalues
of the signals attenuated by the several brain layers whih they have to ross
untilreahing thesalpissolved bytheuse ofampliers. Moderntehnologies
allow on-line ltering of the signals and other ontrols to regulate the signal
output. Furthermore, data displays that followaquisition, oera widerange
of options to represent the data for EEG interpretation. Figures 2.8 and 2.9
illustrate some types of EEG reorders available on the market. Figure 2.8
shows a simple Bluetooth based system with two hannels plus another two
hannels for ground and referene. Figure 2.9 shows a more omplex system
of up to 256 hannels omposed by anEEG ap (A) and EEGamplier(B).
2.3.1 EEG eletrodes
Plaementoftheeletrodeshasbeenstandardisedbyusinga10-20systemthat
usesanatomiallandmarksontheskull. Thenameisbasedontheperentages
usedtodeterminetheeletrodeinstallation. Atotalmeasureisdividedinto10
or 20 perent segments. This system uses the distane between the subjet's
betweenthesubjet'sentraloronalplane,thedistanebetweenbothears,for
lateral measurements asis explained in gure2.10. Nasion is the intersetion
of the frontal bone with the two nasal bones. It an be easily deteted as its
loationis the depressed area between the eyes. Inion is the most prominent
projetion of the protuberaneloatedon the lowerrear of the human skull.
Figure 2.8: Example ofEEG portable systemto reordEEG with 2hannels systemplus
groundandreferenehannels.
Eahsite has aname,aletteridentifyingthelobeandanumbertoidentify
the hemisphere. Even numbers orrespond to the right hemisphere and odd
numberstothelefthemisphere. Thedesignations;F p
(frontopolar),F(frontal),
T (temporal), O (oipital), C (entral), and P (parietal) are utilised in the
1020 system as shown ingure 2.10 [66℄.
Furthermore the letter 'z' makes referene to the entral hannels. For
example, the name C z
orresponds to the position at 50% of the nasio-inion
distane and at50% of thedistane between pre-auriularpoints. Thismeans
it is the exat entre point of the salp. The letter C indiates entral and
'z' makesreferene tothe 0% lateraloset fromthe Centraloronal line.
Modern reording systems provide a ap where the eletrode loations are
already predetermined for ease of use as shown in gure 2.9. These modern
systems require aninreased number of eletrodes and as aonsequene they
are plaed on the 10-10 system, meaning that the distane between them is
(A)
(B)
Figure2.9: Figure(A)showsag.teg.GAMMAsyswith64eletrodesappliedand(B)shows
ag.teg.HIamp amplier[67℄. The apandamplier are interonnetedand linked viaa
USBtoaomputer.
Nowadays the variety of tehnologies used to develop EEG sensors overs
a wide range,from wetand dry eletrodes to wireless EEGsensors. However,
allof them pursue the same objetive: tobe preise.
The termweteletrodes isrelatedtotheneedtouseondutingeletrode
gel to attah it to the salp. The materials used for their onstrution are
several: silver/silver-hloride(Ag/AgCl), tin(Sn), gold(Au)orplatinum (Pt)
[68℄. The Ag/AgCl eletrodes are onsidered the golden standard and they
higher quality signals [69℄. However, the need to use a gel inreases the time
needed to plae the eletrodes and an exess of it may reate shorts between
sensors if it spreadsout.
Figure 2.10: EEG10-20systemeletrodeplaementtoshowhowtheeletrodesshould be
plaedusingperentagevaluesofthesizeoftheperson'ssalp. Referenepointsaremarked
in both views: nasion-inion and A 1
-A 2
. (A) Side-view of a person's headwith the10-20
oordinatesoverlaid. (B)Top-downviewofaperson'sheadwiththeeletrodeoordinates
overlaid[70℄.
By ontrast, dry eletrodes are designed to be eient without the need
for ondutive gel. The absene of gel is substituted by moisture onthe skin,
mainly sweat [69℄. Numerous variations of dry sensors exist on the market.
For instane: stainless steel diss or miro-fabriated silion strutures. This
type of eletrode isused mainlyinresearh as they present some problems of
usability for normal linial appliations due to their instability as they are
muh more diultto seureto the patient than wet eletrodes [69℄.
Thereisanother modelofeletrodethat,ontrarytoawetordryeletrode,
does not require diret physial ontat with the skin. Some examples of
these non-ontat sensors an be found in the literature [71, 69℄. They are
omprised of aset of apaitive eletrodes with a wirelesstransmitter tosend
data to a omputer. These systems have the advantage of being insensitive
to skin onditions and require zero preparation. However, their preisionand
reliabilitystillhave not been proven.
For this researh, the eletrodes used are ative Ag/AgCl ring eletrodes
from g.Te [67℄ implanted in the g.GAMMAap previously mentioned and
2.3.2 Clinialbrain wave bands
Hans Berger was the rst investigator to disover a rhythmi brain wave in
the range of 8-12Hz that he named the
α
band [65℄. Sine his disovery, it has been demonstrated that, irrespetive of the natureof the observed signal,mostofthebrain'sativityhasmultiplefrequeniesthatevolveovertime. The
most important are: the delta, theta, alpha, beta and gamma bands. They
are identied aording to their frequeny and they possess dierent features
that are desribed in table2.2.
Table2.2: SummaryofthemainEEGbrainwavebandsandtheirfeatures[72℄.
Name Frequeny range (Hz) Features Delta (
δ
) 0.5-4Oursin sleep ora vegetative state of
the brain,slowand high amplitude
waves.
Theta (
θ
)4-8
Oursduring lightsleep, quiet foused
meditation. They have been observed
duringmemory retrieval.
Alpha (
α
)8-12
Mediatelevelof onsiousness, relaxed,
awareness of the body, predominant with
losedeyes, prominentabovevisualareas.
Beta (
β
)12-30
Relatedto onsiousness, busy or
anxious thinkingand ative
onentration. Low voltagewaves.
Gamma(
γ
)>30
Withhigh level informationproessing,
forlearning and memory.
2.3.3 Artifats
Reording eletrial ativity from the brain is subjet to non-erebral inter-
ferene due to the high sensitivity of EEG systems. Those soures, named
artifats, an have a non-physiologial origin. For example, eletrial devies
operatingnearbyorphysiologialinterferenesignals originatedfromthe sub-
jet's heart and musle movements. Small movements suh as blinking or
frowning an introdue large spikes in the EEG signals and may deeive the
interpreter tobelievethat theapparent souresare abnormal[73℄. In[74℄, the
authors performed a omparativestudy of the eet of blinking on the signal
tonoiseratio(SNR) forsalpEEGandiEEGsimultaneously. Thisshowsthat
Another soureof noise isthe plaement ofthe eletrodes; if the referenes
to plae the ap or eletrodes are not aurate, the reorded EEG signal will
beaeted by noise. Inthe sameway,if one eletrode isunseure itan move
during the experiment ausing large artifats. In addition to these types of
noise, skin-eletrode noise must be onsidered whih strongly orrelates with
the skin impedane. The eet is redued with the use of the gel in the
ase of wet eletrodes but the issue stillremains under-addressed in the ase
of dry eletrodes, although eorts have been made to negate this eet [69℄.
Reognition and eliminationof the artifats in EEG reordings is an arduous
task, but essential forthe development of pratialsystems [73℄.
In the last deade several methodologies have been proposed to improve
the SNR of EEG measurements, espeially those omprising dierent signal
proessing tehniques designed to redue the noise using a range of temporal
averaging shemes. For example median and weighted averaging, trimmed
estimators, wavelet-based de-noising methods or spatial lters [75℄. In order
to eetively hoose the most appropriate method to deal with noise, several
aspets needtobeonsideredinrelationtothe propertiesof the dataand the
researh questions being asked [76℄.