J. Biomedical Science and Engineering, 2010, 3, 1175-1181
doi:10.4236/jbise.2010.312153 Published Online December 2010 (http://www.SciRP.org/journal/jbise/ JBiSE
).
Published Online December 2010 in SciRes. http://www.scirp.org/journal/JBiSE
Chaotic features analysis of EEG signals during hallucination
tasks of waterloo-stanford standard
Elahe’ Yargholi1, Ali Motie Nasrabadi2
1Biomedical Engineering, Islamic Azad University, Science and Research Branch, Tehran, Iran;
2Biomedical Engineering, Shahed University, Tehran, Iran.
Email: elahe.yargholi@gmail.com; nasrabadi@shahed.ac.ir
Received 9 October 2010; revised 15 October 2010; accepted 20 October, 2010
ABSTRACT
The present study looks carefully at EEG (Electroen-
cephalograph) signals of people after the hypnosis
inductions. Subjects were in three different categories
of hypnotizability based on Waterloo-Stanford crite-
ria; low, medium and high. Signals recorded during
hallucination tasks of Waterloo-Stanford standard
were applied to study the underlying dynamics of
tasks and investigate the influence of hypnosis depth
and concentration on recorded signals. To fulfill this
objective, chaotic methods were employed; Higuchi
dimension and correlation dimension. The results of
the study indicate channels whose chaotic features
are significantly different among people with various
levels of hypnotizability. Moreover, a great consis-
tency exists among channels involved in each task
with brain’s dominant hemisphere and brain lobes’
functions. Another considerable result of the study
was that the medium hypnotizable subjects were
mostly affected by inductions and instructions of the
hypnotizer (more than low or high hypnotizable sub-
jects). The present study demonstrates a remarkable
innovation in the analysis of hypnotic EEG; investi-
gating the EEG signals of the hypnotized as doing
hallucination tasks of Waterloo-Stanford standard
orders.
Keywords: Hypnosis; Hypnotizability;
Waterloo-Stanford Standard; Fractal Dimension
1. INTRODUCTION
To define hypno sis, it could be referred to a mental state
produced through induction. Some people think of that
as a kind of hypnoidal anesthesia while the various neu-
rological researches reveal that hypnosis is a state of
consciousness in which the individual enjoys the high
degree of concentration [1] and as being isolated from
peripheral environment, s/he is extremely suggestible.
For the time being, hypnosis is applied in various
fields such as medicine, psychology, dentistry and …. To
make use of hypnosis as a therapeutic means, the patient
should be appropriately receptive to hypnosis to get the
required depth-point. It is the point at which the patient
would take and accept the therapeutic instructions and
could behave or act based on received inductions so that
the therapy could be efficiently carried out. International
standards are applied to assess/estimate the depth of
hypnosis. According to the international standards, the
individual goes under hypnosis, then the hypnotizer or-
ders her/him to do a special task. How the patient fol-
lows the order helps the hypnotizer to estimate the depth.
This method may cause a hypnosis-depth decrease, so it
is more beneficial to apply methods which estimate the
hypnosis depth ordering the patient and this is what a
great deal of researches aim at.
In recent years due to advances in the field of bio-
medical science, a large number of researches on hypno-
sis and its impact upon different biosignals have been
carried out. While under the influence of hypnosis, the
body experiences physiological shifts such as the change
in the rhythm of the heart, hypotension, resistance of
peripheral vessels, electrical resistance of the skin, basic
metabolism, body temperature, rate and depth of breath-
ing et al. Hypnotic inductions can also change the tone
of muscles and release of endocrine glands so it has led
scholars to do various researches on hypnosis, applying
EEG (Electroencephalograph), fMRI, PET, skin-resis-
tance measurement, heart rate et al. since EEG signal
recording, in comparison with other methods, is more
readily accessible and easier to use, a lot of experts have
applied it to hypnosis investigations.
A wide variety of studies focused on the spectrum of
the hypnotic EEG signals [1-17] while some recent re-
searches were performed not applying the spectrum.
Faber et al. studied the EEG signals during hand rising
in both not hypnotized and hypnotized states [18]. Nas-
E. Yar gholi et al. / J. Biomedical Science and Engineering 3 (2010) 1175-1181
1176
rabadi studied EEG signals in different mental status
(baseline, tasks, hypnosis) in people with different hyp-
notizability [19]. Lee et al. applied fractal analysis to
investigate EEG signals in both states of hypnotized and
not hypnotized [20]. Solhjoo et al extracted fractal di-
mensions of normal and hypnotic EEG signal to classify
different mental tasks [21]. Baghdadi studied features
extracted from hypnotic EEG using improved empirical
mode decomposition (EMD) algorithm [22,23]. Ray
investigated the difference between EEG signals’ fractal
dimensions of lowly and highly hypnotizable subjects
[2]. Behbahani analyzed the nature of hypnosis in right,
left, back and frontal hemisphere in three groups of
hypnotizable subjects by means of fuzzy similarity index
method [24 ,25].
In researches carried out so far, different methods
have been used to examine EEG signals of the hypno-
tized subjects; However, most those researches applied
the EEG signals during induction and investigating the
EEG signals of the hypnotized as doing mental tasks
(standard orders of Waterloo-Stanford) has not still been
realized. Therefore in the pursuit of this purpose, the
current study has been proposed. This study examined
the recorded EEG signals of three groups of people
whose hypnotizability leve ls ranged in low, medium and
high hypnosis receptivity. The signal recording was done
as they were doing the mental tasks (standard orders of
Waterloo-Stanford) so, the impacts of hypnosis depth
and concentration rate on recorded signals of these three
groups of people -with different hypnotizability levels-
could be properly examined and variation of dynamics
throughout di fferent mental tasks would be investigat ed.
The researches performed in the last 25 years indi-
cates the existence of chaotic dynamics in both micro-
scopic (neuron performance) and macroscopic levels
(brain activities during sleep) [26,27], so obtaining the
accurate and better results, through the application of
chaotic methods to EEG investigation in the hypnotized
could be expected. By the means of extracting and com-
paring the chaotic features, the difference between the
recorded EEG throughout the same activities but differ-
ent hypnotizability depth-levels (low, medium and high)
could be examined. Therefore in th is research, the exist-
ing difference of the dominant dynamics in these three
groups with different hypnotizability an d also the kind of
differences have been studied. Should any distinction of
the extracted chaotic features between these three hyp-
notized groups observed, those differences could serve
as the criteria for hypnosis-depth determination to exert
appropriate inductions, EEG examination during hypno-
sis provides the data for studying hypnosis stages and
the transfer from one stage to another.
2. MATERIALS
For the required data in the study, Nasrabadi data base
[19] used. The data were obtained according to 20-10
standard consisting of 19 electrodes. The subjects of the
study included 33 male participants with age range of 32
± 6. The sampling was taken at 256 Hz. The subjects all
featured in left-hemisphere dominancy. Being right-
handed (writing with right hand) was the criterion to
recognize the subjects as left-hemisphere dominants.
The signal recording time was the same for all (16
pm-20 pm). Signal recording were taken twice and in 2
different situations: once in a state of being relaxed with
their eyes closed-baseline signal- and for the second time,
the recording was taken at the state of being hypnotized.
To examine the shifts in the level of hypnosis, EEG sig-
nals are required to be recorded under the state of hyp-
nosis. To do so, following the whole stages of Water-
loo-Stanford standard, a 45-minute audio file was pro-
vided, the same audio file was utilized for hypnosis in-
duction in all the subjects and there was no change in
speech tone. So, all the subjects were placed under the
equal circumstances. The first 15 minutes of audio file
was assigned to the hypnosis induction. It starts includ-
ing an individual- conscious and in normal state, then in
order to determine the hypnotizability score, the partici-
pant is asked to do 12 different tasks as the following
respectively:
1) Hand lowering (ideomotors),
2) Moving hands together (ideomotors),
3) Experience of mosquito (hallucination),
4) Taste experience (hallucination),
5) Arm rigidity (challenge),
6) Dream (memory),
7) Arm immobilization (challenge),
8) Age regressi on (mem ory),
9) Music hallucination ( hall u ci nati on),
10) Negative vi sual (hall ucination),
11) Posthypnotic automatic writing (memory) and
12) Amnesia (memory).
When tasks-performance ends, the scores of hypno-
tizability and the depth of hypnosis are de termined based
on each individual performance [28,29]. The EEG sig-
nals of all electrodes have been recorded throughout
induction and all tasks. In present research EEG signals
of hallucination were chosen to investigate.
3. METHOD
Parameters that represent chaotic behavior may be di-
vided to two categories. The first category indicates dy-
namic behavior. Maximum Lyapunov Exponent (MLE)
is of this category. These parameters state how the sys-
tem behaves oh the nearby trajectories. The second cat-
Copyright © 2010 SciRes. JBiSE
E. Yar gholi et al. / J. Biomedical Science and Engineering 3 (2010) 1175-1181 1177
egory emphasizes the geometric property of basin of
attraction. Fractal dimension is of this category [21]. In
present study two fractal dimensions were used: Higuchi
fractal dimension and correlation dimension.
3.1. Higuchi Algorithm
k new time series are constructed from the signal under
study:
  
,,2,,
for 1,2,,
k
mNm
x
xm xm kxmkxmk
k
mk










(1).
Where m and k indicate the initial time value, and the
discrete time interval between points, respectively. For
each of the k time series k
m
x
, the length of
m
Lk is
computed:
  


111
i
m
xm ikxmikN
Lk Nm
k
k
 



(2).
Where N is the to tal length of the signal x(1), x(2),,
x(N). An average length is computed as the mean of the
k lengths (for m = 1,2,…,k ). This procedure is
repeated for each k ranging from 1 to max , obtaining an
average length for each k. Then the slope of th e best fit-
ted line to the curve of versus

m
Lk k

In Lk
In1 k is
the estimate of Higuchi fractal dimension [30].
3.2. Correlation Dimension
It begins by writing the correlation sum as the following
form:



,1
1
1ij
N
dij
ij
CR Rxx
NN




(3).
Where d is the number of embedding dimension and x
values are vectors in that embedding dimension. Then
the correlation sum tells us the relative number of pair of
points that are located within the distance of R of each
other in this space. is defined to be the number
that satisfies

c
Dd

 
c
D
d
dRCRk
. For
s
at
dd (a satu-
ration value), becomes independent of the embed-
ding dimension d and this is the estimate of correlation
dimension (Figure 1) [31-33]. A d dimensional vector is
the collection of d components
c
D

21
,, ,,
LL L
iiitit id t
xxxxx
 

(4).
Where
L
t is called the time lag and represents the
time interval between the successively sampled values
that we use to construct the vector i
x
. To choose the
time lag
L
t there are two dominant methods. The first
is to choose the lag at which the first zero-crossing of the
Figure 1. as a function of embedding dimension.
c
D
autocorrelation function for the data occurs. Another
method is to select the first local minimum of the aver-
age mutual information function. In present study the
later procedure was applied [34].
4. RESULTS
First, the DC parts of each signal were removed and Hi-
guchi and correlation dimension of both baseline and
hypnotic signals calculated. In the next stage, with the
purpose of normalizing the fractal dimensions of hyp-
notic signals, those of baseline signals were employed.
Applying ANOVA statistic analysis, the following stage
went on in order to find out if there is a significant dif-
ference of their features- either normalized or not nor-
malized- among three hypnotizable groups (low: 1, me-
dium: 2 and high: 3) while going through the same tasks.
Analysis results of hallucination tasks including chan-
nels whose fractal dimension enable us to differentiate
three hypnotizable groups with p-values less than 0.05
and mean values of fractal dimensions for distinguished
groups are presented in the following tables.
As can be seen from Ta bles 1 and 2 of task3 (experi-
ence of mosquito), Tables 3 and 4 of task4 (taste ex-
perience), Tables 5 to 8 of task9 (music hallucination)
and Table s 9 to 12 of task10 (negative visual), in hallu-
cination tasks, channels of left hemisphere (except
task3)and frontal lobe were more efficient and this is
consistent with function of frontal lobe concerned with
the reception and processing of sensory information
from the body [35].
5. CONCLUSIONS
Employing the chaotic methods, the current study was
going to find out the impact, if any, of hypnosis depth on
EEG signals recorded while the individuals were doing
mental tasks under the hypnosis. The results of the study
indicates just a few number of channels, not all, can be
of an aid in discriminating between people with various
levels of hypnotizability and the similarity among those
Copyright © 2010 SciRes. JBiSE
E. Yar gholi et al. / J. Biomedical Science and Engineering 3 (2010) 1175-1181
1178
Table 1. Task3-Higuchi dimension-Normalized.
channels Distinguished groups(mean)
3
F
1a(0.963) 3(1.060)
a. Low: 1, medium: 2, high : 3 .
Table 2. Task3-Correlation dimension-Normalized.
channels Distinguished groups(mean)
4
F
2(5.2483) 3(3.074)
1(5.099) 3(2.051)
4
T
2(4.073) 3(2.051)
Table 3. Task4-Higuchi dimension-Normalized.
channels Distinguished groups(mean)
Z
F
1(0.957) 2(1.009)
1(0.958) 3(1.079)
3
F
2(1.019) 3(1.079)
1(0.961) 3(1.071)
7
F
2(0.998) 3(1.071)
3
C 1(0.981) 3(1.054)
3
P 2(0.995) 3(1.041)
Table 4. Task4-Correlation dimension-Not normalized.
channels Distinguished groups(mean)
4
C 2(8.398) 3(6.758)
Table 5. Task9-Higuchi dimension-Not normalized.
channels Distinguished groups(mean)
3
P 1(0.736) 2(0.858)
channels of the same task type is considerable (Figure
2).
Looking closely at the results, it can be noticed that in
all tasks except task3, channels of left hemisphere were
more efficient and this fact may be due to subjects being
right-handed and left hemisphere dominancy.
A great consistency exists between channels involved
with corresponding brain lobes: task3, task4, task9 and
task10 of hallucination type; channels of frontal lobe.
A remarkable finding yielded through the statistic in
Table 6. Task9-Higuchi dimension-Normalized.
channels Distinguished groups(mean)
1(0.914) 2(1.0199)
Z
F
1(0.914) 3(1.063)
1(0.906) 3(1.092)
3
F
2(1.020) 3(1.092)
1(0.964) 3(1.084)
7
F
2(1.020) 3(1.084)
1(0.972) 3(1.069)
3
C 2(1.013) 3(1.069)
1(0.908) 2(1.015)
1(0.908) 3(1.094)
6
T
2(1.015) 3(1.094)
1(0.875) 2(0.997)
1(0.875) 3(1.056)
3
P
2(0.997) 3(1.056)
Table 7. Task9-Correlation dimension-Not normalized.
channels Distinguished groups(mean)
4
C 1(5.893) 2(8.543)
1(2.275) 2(7.661)
6
T
1(2.275) 3(7.160)
1(4.212) 2(7.628)
3
P
1(4.212) 3(7.590)
2
O 1(3.842) 2(8.967)
Table 8. Task9-Correlation dimension-Normalized.
channels Distinguished groups(mean)
6
T 1(1.562) 2(4.549)
Table 9. Task10-Higuchi dimension-Not normalized.
channels Distinguished groups(mean)
3
P 1(0.715) 2(0.854)
Copyright © 2010 SciRes. JBiSE
E. Yar gholi et al. / J. Biomedical Science and Engineering 3 (2010) 1175-1181
Copyright © 2010 SciRes. JBiSE
1179
Figure 2. channels of hallucination tasks.
Table 10. Task10-Higuchi dimension-Normalized.
channels Distinguished groups(mean)
Z
F
1(0.910) 2(1.004)
1(0.894) 3(1.070)
3
F
2(1.000) 3(1.070)
7
F
1(0.939) 3(1.052)
1(0.850) 2(0.992)
3
P 1(0.850) 3(1.030)
vestigations was that in ex tracting various features of all
10 tasks and all 19 channels the features’ variance of the
medium hypnotizable group was the least and the low
hypnotizable group showed the maximum variance in
each extracted feature. This fact shows that the medium
hypnotizable subjects were mostly affected by induc-
tions and instructions of the hypnotizer (more than low
or high hypnotizable subjects) and the low hypnotizable
subjects had the least affectability while the high hypno-
tizable subjects stood in between, less affectability than
medium hypnotizables and more affectability than low
hypnotizable ones.
For further study these channels could be applied to
proposing various kinds of classifiers so as to define
more objective hypnosis scoring methods.
Moreover, the results of the study can be applied to
study the dynamics of any task by the means of fractal
dimensions.
E. Yar gholi et al. / J. Biomedical Science and Engineering 3 (2010) 1175-1181
1180
Table 11. Task10-Correlation dimension-Not normalized.
channels Distinguished groups(mean)
1(3.666) 2(8.050)
4
F
1(3.666) 3(7.096)
Z
F
1(5.039) 2(8.423)
1(3.510) 2(7.683)
3
F
1(3.510) 3(7.532)
7
F
1(4.763) 2(7.546)
4
C 1(5.117) 2(8.374)
1(4.838) 2(8.534)
Z
C 1(4.838) 3(8.348)
3
C 1(5.277) 2(8.976)
1(4.484) 2(8.260)
Z
P 2(8.260) 3(6.454)
1(3.419) 2(8.842)
3
P 1(3.419) 3(8.022)
2
O 1(3.694) 2(8.608)
1
O 1(5.112) 2(9.186)
Table 12. Task10-Correlation dimension-Normalized.
channels Distinguished groups(mean)
1(1.310) 2(4.068)
3
F
1(1.310) 3(3.586)
3
C 1(2.305) 2(5.431)
Z
P 2(5.077) 3(3.011)
Hallucination tasks:
1) In task3, correlation dimensions of group 3 are less
than others and by contrast higuchi dimension of group 3
is higher than that of group1.
2) In task4, Higuchi dimension of group 3 are greater
than others, Higuchi dimension of group1 is less than
that of group2 and by contrast correlation dimension of
group 2 is higher than that of group3.
3) Task9 indicates group 1 fractal dimensions are less
than those of group 2 and 3 and dimensions of group 3
are greater than those of group 2.
4) Task10 shows group 1 fractal dimensions are less
than those of group 2 and 3 and higuchi dimension of
group 3 is greater than those of group 2 while correlation
dimension of group 3 is less than those of group 2.
Considering the above-mentioned analysis, similari-
ties among dynamics of tasks of the same type are con-
siderable. The results of the study can not be compared
with any other research because there is no previous
study over EEG signal s o f Wate rl oo -Sta nf or d tasks.
REFERENCES
[1] Gruzelier, J.H. (1998) A working model of neurophysi-
ology of hypnotic relaxation. 5th Internet World Con-
gress for biomedical sciences. McMaster University,
Canada, INABIS 98.
http://www.mcmaster.ca/inabis98/woody/gruzelier0814/t
wo.html
[2] Ray, W.J. (1998) Understanding hypnosis and hypnotic
susceptibility from a psychophysiological perspective.
5th Internet Word Congress for Biomedical Sciences,
INABIS98.
http://www.mcmaster.ca/inabis98/woody/ray0556/index.
html
[3] De Pascalisa, V., Ray, W.J., Tranquilloa, I. and D’Amicoa,
D. (1998) EEG activity and heart rate during recall of
emotional events in hypnosis: relationships with Hypno-
tizability and suggestibility. International Journal of
Psychophysiology, 29, 255-275.
[4] Galbraith, G.C., et al. (1970) EEG and Hypnotic Suscep-
tibility. Journal of Comparative and Physiological Psy-
chology, 72, 125-131.
[5] Crawford, H.J. (1998) Brain dynamic shifts during the
elimination of perceived pain and distress: Neuroimaging
studies of hypnotic analgesia. 5th Internet Word Congress
for Biomedical Sciences, INABIS98.
http://www.mcmaster.ca/inabis98/woody/crawford0611/t
wo.html
[6] Crawford, H.J., et al. (1996) Self-Generated happy and
emotions in low and highly hypnotizable person during
waking and hypnosis: Laterality and regional EEG activ-
ity differences. International Journal of Psychophysiol-
ogy, 24, 239-266.
[7] De Pascalis, V. (1998) Brain mechanisms and attentio-
anal processes in hypnosis. 5th Internet Word Congress
for Biomedical Sciences, INABIS98.
[8] De Pascalis, V., et al. (1996) EEG asymmetry and heart
rate during experience of hypnotic analgesia in high and
low hypnotizables. International Journal of Psycho-
physiology, 21, 163-175.
[9] Graffin, N.F., Ray, W.J. and Lundy, R. (1995) EEG con-
comitants of hypnosis and hypnotic susceptibility. Jour-
nal of Abnormal Psychology, 104, 123-131.
[10] Williams, J. D. and Gruzelier, J. (2001) Differentiation of
hypnosis and relaxation by analysis of narrow band theta
and alpha frequencies. International Journal of Clinical
and Experimental Hypnosis, 49, 185-206.
[11] Abootalebi, V. (2000) Investigation of hypnosis on EEG
higher order spectra,” Master Thesis, Sharif University of
Technology.
[12] (2008) Professional Hypnosis DataBank, February.
Copyright © 2010 SciRes. JBiSE
E. Yar gholi et al. / J. Biomedical Science and Engineering 3 (2010) 1175-1181
Copyright © 2010 SciRes. JBiSE
1181
http://www.altor.org
[13] Lubar, J.F., Gordon, D.M., Harrist, R.S., et al. (1991)
EEG correlates of hypnotic susceptibility based upon fast
fourier power spectral analysis. Biofeedback and Self-
Regulation, 16, 75-80.
[14] Brady, B. and Stevens, L. (2000) Binaural-beat induced
theta EEG activity and hypnotic susceptibility, American
Journal of Clinical Hypnosis, 43, 53-69.
[15] Fingelkurts, A.A., Fingelkurts, A.A., Kallio, S. and Rev-
onsuo, A. (2007) Cortex functional connectivity as a
neurophysiological correlates of hypnosis: An EEG case
study. Neuropsychologia, 45, 1452-1462.
[16] Dumas, R.A. (2007) EEG Alpha-Hypnotizability correla-
tions: A review . Psychophysiology, 14, 431-438.
[17] White, D., Ciorciari, J., Carbis, C. and Liley, D. (2009)
EEG correlates of virtual reality hypnosis. International
Journal of Clinical and Experimental Hypnosis, 57,
94-116.
[18] Faber, P.L., Gianotti, L.R.R., Wohlgemuth, P. and Leh-
mann, D. (2000) Frequency domain EEG source loca-
tions during arm levitation under hypnosis: a pilot study.
ISBET 2000, Millennium,Frankfurt on the Main, Ger-
many.
[19] Nasrabadi, A.M. (2003) Quantitative and qualitative
evaluation of consciousness variation and depth of hyp-
nosis through intelligent processing of EEG signals. Ph.D.
Thesis, Amir Kabir University.
[20] Lee, J.S., et al. (2007) Fractal analysis of EEG in hypno-
sis and its relationship with hypnotizability,” Interna-
tional Journal of Clinical and Experimental Hypnosis, 55,
14-31.
[21] Solhjoo, S., Nasrabadi, A.M. and Golpayegani, M.R.H.
(2005) EEG-based mental task classification in hypno-
tized and normal subjects. Engineering in Medicine and
Biology 27th Annual Conference, Shanghai, China, Sep-
tember 1-4, 2041-2043.
[22] Baghdadi, G. (2008) Hypnosis depth determination, using
empirical mode decomposition. Master Thesis, Shahed
University.
[23] Baghdadi, G. and Nasrabadi, A.M. (2009) Estimating
final depth of hypnosis using extracted fractal features by
EMD algorithms. 17th Iranian Conference on Electrical
Engineering. Iran University of Science and Technology,
Tehran, Iran.
[24] Behbahani, S. (2008) Analysis of hypnotic EEG signals
using fuzzy si milarity in dex. Master Thesis, Isl amic Azad
University, Science and Research Branch.
[25] Behbahani, S. and Nasrabadi, A.M. (2009) Application of
fuzzy similarity index method in processing of hypnosis.
Journal of Biomedical Science and Engineering, 2,
359-362.
[26] Aihara, K. and Matsumuto, G. (1986) Chaotic oscilla-
tions and bifurcations in squid giant axon. chaos, Man-
chester University Press.
[27] Babloyantz, A., Salazar, J.M. and Nicolis, C. (1985)
Evidence of chaotic dynamics of brain activity during the
sleep cycle. Physics Letter A, 111, 152-156.
[28] Bowers, K.S. (1998) Waterloo-Stanford group scale of
hypnotic susceptibility, Form C: manual and response
booklet. International Journal of Clinical Hypnosis, 46,
250-268.
[29] Krisch, I., et al. (1998) Experimental scoring for the Wa-
terloo-Stanford group scale. International Journal of
Clinical Hypnosis, 46, 269-279.
[30] Nasrabadi, A.M., Hashemi Golpaygani, M.R., Khalil-
zadeh, M.A. and Sharifi, A. (2003) Comparision between
linear and nonlinear EEG signal processing during dif-
ferent mental activies. Scientific Journal of Amirkabir A,
55, 592-600.
[31] Hillborn, R.C. (2001) Chaos and nonlinear dynamics.
2nd Edition, Oxford University Press, 379-381.
[32] Grassberger, P. and Procaccia, I. (1983) Characterization
of strange attractors. Physical Review Letters, 50,
346-349.
[33] Grassberger, P. and Procaccia, I. (1983) Measuring the
strangeness of strange attractors. Physica D, 9, 189-208.
[34] Fraser, A.M. and Swinney, H.L. (1986) Independent co-
ordinates for strange attractors from mutual information,
Physical Review A, 33, 1134-1140.
[35] Netter, F.H. (1983) The Ciba collection of medical illus-
trations, Volume 1: Nervous System; Part 1: Anatomy
and Physiology. CIBA Limited, West Caldwell.