Neurocomputing 145 (2014) 190200

Contents lists available at ScienceDirect

Neurocomputing
journal homepage: www.elsevier.com/locate/neucom

Letters

Epileptic seizure detection by analyzing EEG signals using

different transformation techniques
Mohammad Zavid Parvez n, Manoranjan Paul
Centre for Research in Complex Systems, School of Computing & Mathematics, Charles Sturt University, Bathurst NSW 2795, Australia

art ic l e i nf o

a b s t r a c t

Article history:
Received 13 August 2013
Received in revised form
4 May 2014
Accepted 24 May 2014
Communicated by Bo Shen
Available online 2 June 2014

Feature extraction and classication are still challenging tasks to detect ictal (i.e., seizure period) and
interictal (i.e., period between seizures) EEG signals for the treatment and precaution of the epileptic
seizure patient due to different stimuli and brain locations. Existing seizure and non-seizure feature
extraction and classication techniques are not good enough for the classication of ictal and interictal
EEG signals considering for their non-abruptness phenomena, inconsistency in different brain locations,
type (general/partial) of seizures, and hospital settings. In this paper we present generic seizure
detection approaches for feature extraction of ictal and interictal signals using various established
transformations and decompositions. We extract a number of statistical features using novel ways from
high frequency coefcients of the transformed/decomposed signals. The least square support vector
machine is applied on the features for classications. Results demonstrate that the proposed methods
outperform the existing state-of-the-art methods in terms of classication accuracy, sensitivity, and
specicity with greater consistence for the large size benchmark dataset in different brain locations.
& 2014 Elsevier B.V. All rights reserved.

Keywords:
EEG
Epilepsy and seizure
Ictal and interictal
EMD
LS-SVM

1. Introduction
The human brain processes sensory information which is
received by external/internal stimuli. Human brain is an organic
electrochemical computer as neurons exploit chemical reaction to
generate electricity [1]. Electroencephalogram (EEG) is a graphical
record of ongoing electrical activity, which measures the changes
of the electrical activity in term of voltage uctuations of the brain
through multiple electrodes placed in the different location of the
brain [2]. In the clinical contexts, the main diagnosis of EEG is to
discover abnormalities of brain activities referred to the epileptic
seizure. A seizure occurs when the neurons generate uncoordinated electrical discharges that spread throughout the brain and
epilepsy is a recurrent seizure disorder caused by abnormal
electrical discharges from brain cells, often in the cerebral cortex
[3]. Another clinical use of EEG is in diagnosis of coma, brain death,
encephalopathies, sleep disorder, etc. Moreover, EEG can be used
in many applications such as emotion recognition [4], video
quality assessment [5], alcoholic consumption measurement [6],
sleep stage detection [7], change the brainwaves by smoking [8],
mobile phone usages [9], etc.
People may experience abnormal activities in sensation, movement, awareness, and behavior during seizure as a result they

Corresponding author.
E-mail address: mparvez@csu.edu.au (M.Z. Parvez).

http://dx.doi.org/10.1016/j.neucom.2014.05.044
0925-2312/& 2014 Elsevier B.V. All rights reserved.

cannot perform normal task. Ictal and interictal both are medical
condition of seizure where ictal represents the period of seizure
and interictal represents the intermediate period between two
seizures. Note that interictal is signicantly different from normal
non-seizure (mainly for healthy people) signal in terms of signal
characteristics. A prediction of seizures (i.e., ictal) from interictal
could aware a patient to put away from the next seizure and also
can make sure better treatment and precaution. A number of
existing EEG features extraction and classication techniques
[1117] are able to classify seizure and non-seizure EEG signals
for a small size dataset almost perfectly. However, those methods
are struggling to provide acceptable classication accuracy for ictal
(i.e., seizure period) and interictal (i.e., period between seizures)
EEG signals due to the non-abruptness and inconsistence phenomena [10] of the signals in different brain locations. Moreover, it is
more challenging to get acceptable label of accuracy by a particular
method if we do not have any domain knowledge of the available
EEG signals due to different types of seizures (i.e., generalized or
partial), patients, hospital setting, brain location, and artifact. Note
that ictal signals are considered the EEG signals during the seizure
period when a patient shows abnormal activities, whereas interictal signals are considered as non-ictal signals between two
seizure periods of an epileptic patient. Thus, we can consider the
characteristics of interictal signal as a middle stage of non-seizure
and ictal signals (although a patient may show normal brain
activities similar to the non-seizure signals during the interictal
period).

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

191

Fig. 1. Ictal and interictal EEG signals from different datasets and different brain locations: (a) Seizure and non-seizure EEG signals from the small dataset [19], the rst row
and remaining four rows indicated the seizure and non-seizure signals respectively [19]. (b) Ictal (patient one, 7th datablock, channel one) and interictal (patient one, 95th
datablock, channel one) of Frontal lobe signals [22] and (c) Ictal (patient 12, 15th datablock, and channel one) and interictal (patient 12, 35th datablock, and channel one) of
Temporal lobe signals [22].

Existing methods [1118] used small epilepsy dataset [19] (for

detailed information of the dataset, see Section 3.1). The small
dataset with duration 23.6 s has seizure (i.e., ictal) and non-seizure
signals which can be distinguished by their visual phenomena
such as magnitude of amplitude and changing rate of frequency
(see in Fig. 1(a)). For the non-seizure signal the amplitude is low
and the frequency is high while the nature of seizure signal is
totally opposite (see Fig. 1(a)). Fig. 1(b) and (c) shows the ictal and
interictal EEG signals from a large dataset [22]. The gure
demonstrates the non-abrupt phenomena (i.e., not easily distinguishable between ictal and interictal based on amplitude and
frequency) of the ictal and interictal signal for both cases of Frontal
and Temporal lobes compared to the dataset in [19]. Moreover,
EEG signals from different locations exhibit different phenomenal
activities for an ictal and interictal period. Thus, the classication
of ictal and interictal EEG signals is challenging compared to that
of seizure (i.e., ictal) and non-seizure EEG signals. It also demonstrates that the characteristics of EEG signals are varied from
dataset to dataset.
Birvinskas et al. [17] used DCT (discrete cosine transformation) to
reduce the EEG signals and then extracted features from the
reduced signals to classify EEG signals. They achieved good
classication accuracy by keeping low frequency DCT coefcient
for feature extractions. Worrell et al. [33] claimed that the seizure
onset frequencies are in the range of gamma frequency i.e., high
frequency band (typically between 30 to 100 Hz). They explicitly
mentioned that seizure onset frequencies are on the range of
frequency 74.3 76 13.7 and 23.7 76 2.8 Hz for Frontal and
Temporal lobes, respectively. They also observed that dominant
spectral peaks are observed in the high frequency spectral components for the Frontal lobe and in the lower gamma frequency for
the Temporal lobe during the seizure period. Other literature such

as Panda et al. [11], Ocak et al. [13], and Ghosh-Dastidar et al. [12]
used other frequency bands such as delta, theta, alpha, and beta
with gamma band to extract features for EEG signal classication.
From the above ndings and the characteristics of the different
datasets, we nd that the frequency range of seizure onset can be
varied based on the brain location, seizure type (general/partial),
patient, and hospital setting (e.g., capturing machine quality,
artifacts, etc.). However, we observe that the signal has larger
spikes i.e., high amplitude in the ictal (i.e., seizure) period
compared to the interictal period. Thus, based on the temporal
and spatial features of raw EEG signals and the above ndings of
the recent literatures, our hypothesis is that the distinguishing
features between ictal and interictal signals would be in the high
frequency spectral area irrespective of the brain location, patient,
type of seizure, and hospital setting.
To verify our hypothesis we perform an experiment to nd the
gamma frequency of signals which provides the maximum amplitude using ictal and interictal signals in different brain locations.
The results for 200 EEG signals from ictal and interictal are shown
in Fig. 2. The gure shows that relatively high gamma frequencies,
i.e. high frequencies in the signals provide the maximum amplitude of ictal compared to the interictal signals for both Frontal and
Temporal lobes from different patients. We use the maximum
amplitude criteria for the gamma band selection in this analysis as
our observation and Worrell et al.'s observation indicate that
dominant spectral peaks for the seizure onset are in the high
frequency (i.e., gamma band). Thus, the features extracted from
gamma frequencies, i.e. the high frequencies should provide better
classication accuracy between the ictal and interictal irrespective
of brain location, patients, type of seizure, and hospital setting.
Based on the above mentioned observation, in this paper we
proposed generic seizure detection techniques which use different

192

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

Fig. 2. Gamma frequencies which provide the maximum amplitude for ictal and interictal signals (from a standard dataset [22])

established transformations and decompositions to get different

statistical features from the high frequencies component of the
signals to classify the ictal and interictal signals using the least
square support vector machine (LS-SVM) [23]. Although we use
a number of established transformation and decompositions, we
use weighted high frequency coefcients to extract the features. In
the experiment we investigate the strength of each transformation
and decomposition for the capability of classication accuracy as
well as the combined strength through different feature extractions. We use a large dataset [22] consists of raw ictal and
interictal signals (no explicit artifact removal technique is applied)
from different brain locations, different seizures, and patients to
verify the performance of the proposed methods against two
state-of-the-art methods. The experimental results demonstrate
that the proposed generic methods outperform the existing
methods in terms of accuracy (i.e., overall), sensitivity (i.e., ictal
or positive), and specicity (i.e., interictal or negative) with greater
consistence for the challenging ictal and interictal EEG signals
classication.
The contributions and novelties of the proposed methods are in
the capacity to (i) classify challenging ictal and interictal signals
irrespective of brain location, patient, type of seizure, and hospital
setting with a certain range of artifact tolerance, (ii) identify that
high frequency component of EEG signals has the distinguishing
features to classify ictal and interictal EEG signals, (iii) formulate
a weighing vector for high frequency coefcients to provide
individual importance of different coefcients, (iv) provide the
rationality and way of extracting high frequency components from
each transformation/decomposition technique, (v) to analysis EEG
signals for seizure detection, best of our knowledge we are the rst
one to use two dimensional SVD, (vi) identify the distinguishing
features (provided pictorial comparison) for ictal and interictal
signals for classication, and (vii) provide comprehensive result
analysis using receiver operating characteristics (ROC) curve, Whisker-box, graphical view of classication outcome, and classication performance in terms of sensitivity, specicity, and accuracy.

2. Existing methods for seizure detection

Existing feature extraction and classication methods based on
wavelet [1113] and Fourier transformation [14] were employed
for the detection of seizure in EEG signals. Panda et al. [11]
computed various features, such as energy, entropy, and standard
deviation (STD) using discrete wavelet transformation (DWT) and
used support vector machine (SVM) as a classier. Dastidar et al.
[12] applied wavelet transformation to decompose the EEG signals
into different range of frequencies and then extracted three

features, such as STD, correlation dimension, and the largest

Lyapunov exponent (quantifying the non-linear chaotic dynamics
of the signals) and applied different methods for classication.
Ocak [13] proposed a fourth level wavelet packet decomposition
method to decompose the normal and epileptic EEG epochs to
various frequency-bands and then used genetic algorithm to nd
optimal feature subsets which maximize the classication performance. Polat et al. [14] extracted feature using fast Fourier transform (FFT) and then classied using decision making classier. The
techniques [11,12,14] have used small dataset [19] (for detailed
information of the dataset, see Section 2.1). Sample examples of
the dataset are shown in Fig. 1(a) The classication accuracies
were 91.20% [11], 96.70% [12], and 98.72% [14], respectively. Ocak's
[13] experimental dataset was of the same pattern with the
dataset in [19] and the classication accuracy was 98%.
Other feature extraction and classication methods [1518]
based on empirical mode decomposition (EMD), principle component
analysis (PCA), and genetic algorithm (GA) have been proposed.
Liang et al. [15] extracted EEG features where the approximate
entropy(ApEn) analysis was evaluated for its ability to analyze
multiclass EEGs, and the performance of ApEn analysis was
enhanced by incorporating the power of EEG sub-bands or autoregressive models. PCA and GA were compared for their ability to
select features by applying various linear and nonlinear methods
for classication. The maximum accuracy was 98.67%. Pachori [16]
decomposed EEG signals into intrinsic mode function (IMF) using
EMD and then computed mean frequency for each IMF by Fourier
Bessel expansion [20] to differentiate seizure and non-seizure EEG
signals. Zhang et al. [21] proposed an approach for the pattern
recognition of four complicated hand activities, such as grasping
a football, a small bar, a cylinder, and a hard paper, based on EEG
signal, in which each piece of raw data sequence for EEG signal is
decomposed by wavelet transform. After forming a matrix they
applied singular value decomposition (SVD) on the matrix to nd
the singular value feature vectors. These features are used as input
to articial neural network to discriminate four hand activities and
they got 89.00% classication accuracy.
Recently, Bajaj et al. [18] extracted amplitude modulation and
frequency modulation bandwidth as features from IMF using EMD.
Among the existing contemporary techniques, Bajaj et al.'s technique is the latest and the best in terms of performance. They used
the LS-SVM [23] technique for the classication of seizure and
non-seizure EEG signals using the small dataset [19]. They
obtained 98.0 to 99.5% accuracy using radial basis function (RBF)
kernel and also obtained 99.5 to 100% accuracy using Morlet
kernel.
We have observed that the existing methods [1118] exhibit
almost perfect classication accuracy for the small dataset of

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

seizure and non-seizure EEG signals. However, their performance

is below the expected results using the large dataset [22] for ictal
and interictal EEG signal classications. For example, when we
applied the Bajaj et al.'s [18] feature extraction and classication
method to differentiate ictal and interictal EEG signals for the rst
IMF in the large dataset from Epilepsy Center of the University
Hospital of Freiburg [22], the classication accuracy is 77.90% for
Frontal lobe and 80.20% for Temporal lobe which is far below the
accuracy 98.50% obtained using the small dataset [19] of seizure
and non-seizure EEG signals.
The paper is organized as follows: the details of the proposed
four methods with dataset, feature extractions, classication
techniques, and computational time are described in Section 3;
the detail experimental results are explained in Section 4 while
Section 5 concludes the paper.

3. Proposed methods
In this paper, we propose four methods based on the various new
features extracted from DCT, DCTDWT, SVD, and IMF. We use the
transformation and decompositions to separate high frequency
components and then apply innovative ways to extract different
statistical features. When we extract features from the high frequency component (i.e., IMF and detail signals respective) using EMD
and DWT, we take total length of the high frequency components
and we do not use any weighting for them as all the coefcients
represent the high frequency components. As SVD and DCT provide
coefcients of all frequencies, we need to select a subset of the
coefcients which can represent high frequency components. We
investigate the strength of each transformation and decomposition
for the capability of classication accuracy as well as the combined
strength through different feature extractions. The dataset, individual
method, their corresponding rationality, and detailed methodology of
feature extractions are described in the following sub-sections:
3.1. Dataset
Small Dataset [19]: The existing seizure and non-seizure feature
extraction and classication schemes [1118] except [13] used
small dataset [19]. The dataset consists of ve subsets, each subset
contains 100 single-channels recoding, and each recoding has
23.6 s in duration captured by the international 1020 electrode
placement scheme, i.e. 32 electrodes [14]. Among them two
subsets are taken from healthy volunteers, two subsets are taken
from seizure free intervals and another subset is also taken during

193

seizure period [18] (see sample examples in Fig. 1(a)). Ocak [13]
used different datasets with the same pattern of small dataset
where 300 were normal and 100 were epileptic patients.
Large Dataset [22]: Seizure and brain discharges are much more
complicated than what can be seen in the scalp. Thus, in the paper we
have used intracranial EEG dataset using grid, strip and depth electrodes
rather than scalp EEG. The dataset used in the paper was recorded at
Epilepsy Centre of the University Hospital of Freiburg, Germany [22]. The
database contains invasive EEG recordings of 21 patients suffering from
medically intractable focal epilepsy. The data obtained by Neurole NT
digital video EEG system with 128 channels, 256 Hz sampling rate, and
16 bit analog-to-digital converter. According to the dataset [22], epileptic
EEG signals can be classied into (i) ictal (i.e., seizure), (ii) preictal (a
certain period before seizure onset), (iii) interictal (period between
seizures), and (iv) non-seizure (period without any seizure activity after
a certain time of interictal). Normally, duration of ictal is varied from few
seconds to 2 min. The ictal records contain epileptic seizures with at
least 50 min of preictal data preceding each seizure. The median time
period between the last seizure and the interictal data set is 5 h and
18 min, and the median time period between the interictal data set and
the rst following seizure is 9 h and 36 min [29]. As suggested in the
dataset, we have treated preictal and ictal period as an ictal signals in
our experiments. We do not consider any non-seizure signals into the
interictal period as we carefully select the same duration of interictal
signal (with the ictal signal) from the above duration of interictal signal.

3.2. Feature extraction using DCT

DCT is a transformation method for converting a time series
signal into basic frequency in such a way that the DCT coefcients
are arranged from low frequency to high frequency components.
Low frequency components represent the coarse signals and high
frequency components represent the detail signals. DCT is widely
used in the image processing and video coding areas to compress
image signals based on their frequencies [34]. As the ictal and
interictal EEG signals have different amplitudes and frequencies
(not visually separable), thus the most distinguishable features
should be located in the high frequency components of DCT
coefcients see in Fig. 3(a). As the EEG signal is non-linear and
non-stationary in nature, thus, for real time processing of EEG
signals, DCT may not be correct to directly correspond to the
frequency analysis, however, if we segment the EEG signals in time
window and apply DCT on them to nd DCT coefcients; we can
avoid non-linear and non-stationary nature of the signals. In our
experiment, we apply DCT on the benchmark dataset [22] and take

Fig. 3. (a)High frequency DCT coefcients and their energy and entropy for the signals of Temporal lobe; the rst two rows show the high frequency DCT coefcients (i.e.,
25%) of ictal and interictal signals, while the last row shows the energy and entropy of using the high frequency DCT coefcients (i.e., 25%). (b) High frequency DWT
coefcients and STDs of high frequency DCT and DWT coefcients for the EEG signals of Temporal lobe; the rst two rows show that the high frequency DWT coefcients of
ictal and interictal signals, while the last row shows the STD of the high frequency DCT and DWT coefcients of ictal and interictal EEG signals.

194

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

the higher frequency component (i.e., the last quartile (Q4) i.e., last
25% of DCT coefcient). Then, we extract statistical features such as
entropy and energy of the weighted higher frequency coefcients to
classify ictal and interictal signals.
The extraction of the statistical features from the Q4 is not
straightforward. We use heuristics to provide different weight to
the different coefcients and investigate the classication accuracy. Although we try different type of weighting approximations,
the experimental results show that exponential weight performs
well. The weight wi for the i-th position coefcient is selected as
1
. We also try
wi e  xi where xi  xj Number of Selected
Coef f icients
different and test for the classication results. We nd that
2 provides the best results for the dataset. Fig. 4 shows an
example of weights. According to the proposed weight, we provide
more weight to the relatively low frequency coefcients and less
weight to the relatively high frequency coefcients within the
higher frequency area of the overall signal as sometimes the high
frequency coefcients in the tail would be very small.
We dene weighted entropy and energy as follows:
n

n

2
Energy
W i X i t
and Entropy  W i Y 2i tlnY 2i t
i1

i1

where n is the length of signal, X(t) is the higher frequency DCT

coefcients and Y(t) is the normalized value of X(t).
The representation of ictal and interictal data for different higher
frequency DCT coefcients and their entropy and energy are given in
Fig. 3(a). It can be observed that the changing rate of frequencies is
higher for interictal signal than ictal signal from Temporal lobe. It is
also interesting to observe that the amplitude of the higher frequency
DCT coefcients for ictal signals is larger compared to that of interictal
signals. Note that although the original ictal and interictal EEG signals
are not visually separable, the high frequency DCT coefcients of ictal
and interictal signals are visually separable. The third row of Fig. 3
(a) also demonstrates that the weighted energy and entropy extracted
from the higher frequency DCT components have different values for
ictal and interictal signals. Thus, weighted entropy and energy would
be effective features for ictal and interictal signals classication.
3.3. Feature extraction using DCTDWT
DWT is a transformation and widely used in the image and signal
processing research to decompose the image/signals into different
frequency and bands. DWT can be specied by low pass (LP) and high
pass (HP) lters named as standard quadrature mirror lters [28]. The
cut-off frequency of these lters is one-fourth of the sampling
frequency. The bandwidth of the lter outputs is half the bandwidth
of the original signal, which generates downsampled output signals
without losing any information according to the Nyquist theorem [13].

The down-sampled signals from LP and HP lters refer to rst-level

approximation and details of the original signal respectively [24]. DWT
is sometimes more effective compared to DCT to separate signals into
low and high frequency components when the signal is longer. Thus,
we also investigate the effectiveness of the DWT to extract some other
statistical features (such as STD) to classify ictal and interictal signals.
In our second approach, we apply single-level one dimensional DWT
on the EEG signal to nd the high frequency components from the
signal. Then, we calculate the STD from the detail frequency (i.e., high
frequency) components of the DWT coefcients and use as a feature.
We use the whole high frequency component for feature extraction as
DWT separates the signals into low and high frequency components.
Another feature is extracted using the STD of the weighted high
frequency DCT coefcients (similar to Fig. 4). Two STD features from
DCT and DWT higher frequency components are then used for the
classication purpose. Note that we use different primary functions of
the wavelet transformation including Daubechies (db) series. We nd
that db16 is the best for our experiment.
In the experiment we take only the high frequency DWT components. The rst two rows in Fig. 3(b) show the higher change of
frequency and lower amplitude for interictal signal compared to ictal
signal from Temporal lobe. The third row of Fig. 3(b) represents the
STDs generated from the high frequency DCT coefcients (weighted)
and DWT for ictal, which are higher than that of interictal signal. Thus,
the STDs extracted from the high frequency DCT (weighted) and DWT
coefcients can be good features for classication.
3.4. Feature extraction using SVD
SVD decomposes two dimensional signals into three matrixes in
such a way that the middle matrix (i.e., singular value) has unique
properties to separate signals. The middle matrix, containing the
square root of eigenvalues and diagonalized, contains much less nonzero coefcients compared with the coefcient matrices of other
transformations [35]. The values of diagonal matrix are with descending order from top-left to bottom-right. As the original ictal and
interictal signals are almost of the same nature, we do not expect a
signicant different values of other two matrixes after SVD, however,
we expect some distinguishable values on the extreme part (i.e.,
bottom-right) of diagonal matrix. Thus, any feature extracted from
the bottom-right singular values could be used to classify ictal and
interictal signals. To the best of our knowledge we are the rst to use
two dimensional SVD to analyze EEG signals for seizure detection.
To extract the features using SVD, we reshape the EEG signals
into a two dimensional matrix. The column vector of EEG signal is
rearranged to square matrix for the computation of singular value
components using SVD. From the singular values, we take the
weighted Q4 (weight is similar to Fig. 4 but the number of selected
coefcients are different) of the non-zero diagonal values to
calculate the STDs and energy and treat as the two features to
classify ictal and interictal signals. SVD allows transforming any
given matrix A A C mxn to diagonal form using unitary matrices, i.e.
A UV H where Amn is a matrix, U mm and V nn are orthogonal
matrix, and mn matrix with non-negative diagonal entries.
The columns of U are called left singular vectors and the rows
of VH are called right singular vectors. If A is a rectangular m  n
matrix of rank K then U will be m  m and will be
2
3
1 0 0
60
07
2
6
7
6
7
6 7
7
6
2
60 0 7
n7
6
6
7
4
5
0

Fig. 4. Weight for different coefcients of selected high frequency coefcients.

where 1Z 2 Z n-1 Zn.

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

195

Fig. 5. (a) Non-zero singular value of SVD for Frontal lobe. The rst row shows that the non-zero singular value distribution for ictal and interictal signals from Frontal lobe
and the last row shows the STD and energy of the last 25% non-zero singular values of ictal and interictal EEG signals. (b) The rst IMFs of ictal and interictal signals with the
extracted features; the top two rows represent the rst IMF from the ictal and interictal signals of Temporal lobe; bottom row represents STD of the rst IMF and STD of the
normalized PSD (power spectral density) of the same IMF of ictal and interictal EEG signals.

The data distribution of non-zero singular values of the ictal

and interictal signals from the Frontal lobe is shown in Fig. 5(a).
The top part of the gure clearly shows that the Q4 singular values
of interictal signals are larger than that of the ictal signals. The
bottom part of Fig. 5(a) shows the STD and energy extracted from
the weighted Q4 of non-zero singular values from ictal and
interictal signals respectively. The gure exposes that the STDs
extracted from the ictal and interictal signals have distinguishable
values, thus they can be used for ictal and interictal signal
classication.
3.5. Feature extraction of IMF using EMD
The main strength of EMD is its ability to analyze nonlinear and
non-stationary signals like EEG signals. When we apply EMD on
a signal, the signal is decomposed into a nite number of signals
into IMFs, which are ordered from higher frequency component to
lower frequency component. The number of IMFs in a signal
depends on the local characteristics of the signal rather than
pre-dened number of IMFs. This feature makes each IMF crucial
to identify some characteristics of the original signal. Since the
IMFs are the representations of different frequencies/amplitudes
of the original signal and the distinguishing features between ictal
and interictal EEG signals lie on the frequency/amplitude, our
hypothesis is that any feature extracted from IMFs will be an
effective feature to classify ictal and interictal successfully. We also
observe that the ictal and interictal signals are deferred in higher
order frequency components, thus our other hypothesis is that
features extracted from the rst few IMFs should be enough to
classify ictal and interictal signals successfully. Through the
decomposition we keep only few mode functions of the original
signals for the nal feature extraction, however, we are able to
keep the important mode functions for the classication purpose.
Experimental results reveal that our both hypothesis are correct as
extracted features from the rst IMF successfully classify ictal and
interictal signals (see Section 4). Thus, IMFs can be used in
classication applications where the original signals are nonlinear and non-stationary and distinguishing features exist on
the different frequencies/amplitudes.
Each IMF satises two following conditions: (i) the number of
extrema and the number of zero crossing are identical or differ at

most by one and (ii) the mean value between the upper and the
lower envelope is equal to zero at any time.
The EMD algorithm can be summarized as follows:
1. Extract the extrema (minima and maxima) of the signal xt:
2. Interpolate between minima and maxima to obtain
min t and max t:
3. Calculate local mean mt min t max t =2.
4. Extract the detail dt xt  mt.
5. Check dt is an IMF according to the conditions which are
mentioned above. If yes, repeat the procedure from the step
1 on the residual signal rt xt  mt. If no, replace xt with
dt and repeat the procedure from step 1.
The pictorial scenario of generating IMFs is shown in Fig. 6. The
signal xtcan be represented as a combination of IMFs and
residual component:
L

xt di t r L t

i1

where di and rL are the i-th IMF and L-th is residual signal.
We observe that features extracted from the rst two IMFs
provide the best classication results. To reduce computational
time, we consider only the rst IMF from the EMD and then
calculate STD of IMF to capture the variation of the signal. By
ignoring other IMFs, we do not sacrice any signicant classication
accuracy. The strength of a signal is distributed in the frequency
domain, relative to the strengths of other ambient signals, is central
to the design of any lter intended to extract or suppress the signal.
To nd the strength of the signal, we apply power spectral density
(PSD) on the rst IMF. Then, determine the STD feature from
normalized PSD. The calculation of PSD is as follows:

2
1

Pwi
Dwi

N

where Dwi is the discrete Fourier transform of IMFdi t: Normalized

PSD can be dened as
pi Pwi = Pwi
i

196

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

Fig. 6. Extracted IMFs by EMD from Temporal lobe (Patient 12, channel one, 15th datablock for ictal, and 35th datablock for interictal).

Table 1
Extracted features of proposed and existing methods.
Proposed methods

Feature-1

Q4WHF_SVD (Quartile four weighted high frequency component using SVD )

STD on 25% weighted high frequency Energy on 25% weighted high frequency
singular values
singular values
Energy on 25% weighted high
Entropy on 25% weighted high frequency DCT
frequency DCT coefcients
coefcients
STD on the rst IMF (i.e., high
STD on normalized PSD
frequency components)
STD on 25% weighted high frequency STD on detail signals (i.e., high frequency
DCT coefcients
components) after apply DWT
Combined all above features
Feature-1
Feature-2
Amplitude modulation bandwidth
Frequency modulation bandwidth
Features from 6% of quantized DCT coefcients at lower frequency

Q4WHF_DCT (Quartile four weighted high frequency component using DCT )

FIMF_EMD (First IMF using EMD )
Q4WHF_DCT HF_DWT (Quartile four weighted high frequency component
using DCT and high frequency from DWT)
Combined
Existing methods
EMD[18]
DCT [17]

In the experiment, we extract two features such as STD of the

rst IMF and STD of normalized PSD of the rst IMF. From the rst
two rows in Fig. 5(b), it can be observed that the rst IMF of ictal
signal carries the higher amplitude compared to the interictal
signal.
The third row of the gure represents the STD of the rst IMF
and STD of PSD of the same IMF from using ictal and interictal
signals. The STDs for both cases of interictal signal are signicantly
larger than that of ictal signal. Thus, the two new features
extracted by EMD method can be good features for classication.
Table 1 presents the summary of all corresponding features use
for ictal and interictal classication by each method.
3.6. Classier
We extract a number of features from different transformations/decompositions. To classify ictal and interictal signals, we
need to use a classier. The goal of a classier is to nd patients
states such as ictal (class 1) and interictal (class 2) using machine
learning approaches with cross-validation. The challenge is to nd
the mapping that generalizes from training sets to unseen test
sets. For the cross-validation, data are partitioned into training set
and test set. In our experiment we use 4-fold cross validation
where each dataset was randomly divided into 4 splits, where
three of them were used for training and the fourth one used for
testing.
To classify the ictal and interictal signals, we extract various
features from DCT, DWT, SVD, and IMF of EMD. For classication, we

Feature-2

use SVM-based classier as the SVM is one of the best classiers in

the EEG signal analysis. SVM is a potential methodology for solving
problem in linear and nonlinear classication, function estimation,
and kernel based learning methods [25]. It can minimize the
operational error and maximize the margin hyperplane, as a result
it will maximize the classication performance [25]. LS-SVM is the
extended version of SVM and it is closely related to regularization
networks and Gaussian process, and it also has primal-dual interpretations [23]. The major drawback of SVM is in its higher
computational burden of the constrained optimization programming. However, LS-SVM can solve this problem [26].
The classier is a LS-SVM, which learns nonlinear mapping
from the training set features {x}i 1nT, where nT is the number
of training features into the patient's state, ictal ( 1) and interictal
( 1). For unbiased classication results [27], we divide the whole
1000 trials into four subsets; in each subset we randomly select
75% samples for the validation training set and the remaining 25%
for the validation testing set. Let fyi gi C 1;2 designate the LS-SVM
validation test outputs mapping to class 1 or class 2. The equation
of LS-SVM can be dened in [18] as
"

f x sign i yi Kx; xi c

#
6

i1

where K(x, xi) is a kernel function, i are the Lagrange multipliers, c

is the bias term, xi is the training input, and yi is the training
output pairs. Linear, RBF and Morlet kernel are used in our
experiments and RBF and Morlet functions can be dened in [18]

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

as
kx; xi exp  x  xi =2s
2

where s controls the width of the RBF function

h
i
d
kx; xi cos 0 xk xki =a exp  xk  xki =2a2

k1

where d is the dimension, a is the exible coefcient, and xki is the

k-th component of i-th training data. Matlab codes for LS-SVM
classier are available at http://www.esat.kuleuven.ac.be/sista/
lssvmlab/.
3.7. Computational time
Feature extractions using IMF through EMD take longer time
due to two-fold iterations to nd all IMFs. In the rst iteration
EMD decomposes a signal and checks whether the signal makes an
IMF and in the second iteration it subtracts the current IMF from
the signal and continues for the next IMF. Even if we stop the
iteration for the rst IMF, the EMD-based technique takes a full
rst iteration for calculating the IMF. The experimental data
reveals that the proposed technique using DCT, DWT, and SVD
takes less computational time compared to the EMD-based technique as DCT, DWT, and SVD transformation is faster compared to
the EMD decomposition. Thus, the proposed methods (except
FIMF_EMD method, i.e. First IMF using EMD) take insignicant
time compared to the existing method [18].

4. Experimental results
Sometimes EEG signals have line noise and other kinds of
artifacts due to muscle and body movements. Notch lter and
independent component analysis (ICA)-based method are recommended to remove the line noise and artifacts respectively for

197

sufcient amount of EEG signals [3032]. However, in this experiment we do not use any explicit noise/artifacts removal technique
to see the tolerance of the proposed method against noise/
artifacts.
We propose four methods and the experimental results of the
proposed methods are compared with existing methods [17,18]
and corresponding features are presented in Table 1.Our ultimate
target is to classify ictal and interictal EEG signals using LS-SVM
with different kernels such as RBF and Morlet where the values of
regularization parameter and kernel parameter are 10 and 6,
respectively for both kernels. We also use linear kernel. In this
paper, four methods are proposed based on the new features of
ictal and interictal EEG signals captured from Frontal and Temporal
lobes using DCT, DWT, SVD, and IMF of EMD to see the strength of
individual transformation/decomposition techniques for seizure
detection. Then we also test the classication accuracy by combining all features. In our experiment, we use 200 signals from ictal
and 800 signals from interictal EEG signals. After training and
testing of all features, we calculate average value from the four
subsets to compute sensitivity, specicity, and accuracy [18]. The
sensitivity, specicity, and accuracy are dened as
Sensitivity TP=TP FN  100

Specif icity TN=TN FP  100

10

Accuracy TP TN=TP TN FP FN  100

11

where TP and TN represent the total number of detected true

positive events and true negative events respectively. The FP and
FN represent false positive and false negative respectively.
The technique in [18] claims that the second IMF provides
better classication results while they use frequency and amplitude modulation bandwidth features for the dataset [19]. We
conduct experiments using rst four IMFs; however, we provide
results using only the rst IMF presented in Table 2 as it carries the

Table 2
Sensitivity, specicity, and accuracy for different features of ictal and interictal EEG signals from Frontal and Temporal lobes.
Classier

Brain location

Kernel

Lobe

RBF

Frontal

Temporal

Average

Morlet

Frontal

Temporal

Average

Linear

Frontal

Temporal

Average

Criteria

SEN
SPE
ACC
SEN
SPE
ACC
SEN
SPE
ACC
SEN
SPE
ACC
SEN
SPE
ACC
SEN
SPE
ACC
SEN
SPE
ACC
SEN
SPE
ACC
SEN
SPE
ACC

Existing method

Proposed method

EMD[18]

DCT[17]

Q4WHF_SVD

Q4WHF_DCT

FIMF_EMD

Q4WHF_DCT HF_DWT

Combined

29.41
80.51
77.90
66.67
80.28
80.20
48.04
80.40
79.05
33.82
81.01
77.80
84.62
83.54
83.60
59.22
82.28
80.70
28.95
80.35
78.40
84.21
85.28
85.20
56.58
82.82
81.80

8.70
79.73
78.10
4.00
98.63
79.70
6.35
89.18
78.90
0.00
79.78
78.9
2.00
99.00
79.60
1.00
89.39
79.25
27.28
80.61
76.6
6.00
96.13
78.10
16.64
88.37
77.35

66.66
82.20
81.50
100
83.51
84.20
83.33
82.86
82.85
63.83
82.16
81.30
100
84.30
85.10
81.92
83.23
83.20
80.95
81.31
81.30
98.00
84.11
84.80
89.48
82.71
83.05

100
81.30
81.60
98.24
84.73
85.50
99.12
83.02
83.55
88.89
81.26
81.40
98.41
85.27
86.10
93.65
83.27
83.75
100
80.89
81.10
97.22
82.88
83.40
98.61
81.89
82.25

95.83
81.86
82.20
97.10
85.71
86.50
96.47
83.79
84.35
93.33
81.12
81.30
90.12
86.18
86.50
91.73
83.65
83.90
100
81.55
81.90
98.11
84.37
85.10
99.06
82.96
83.50

76.78
83.37
83.00
97.47
86.65
87.50
87.13
85.01
85.25
60.00
83.96
81.80
93.68
86.32
86.90
76.84
85.14
84.35
100
80.81
81.00
98.00
84.12
84.80
99.00
82.47
82.90

49.14
83.82
79.80
96.47
87.10
87.90
72.81
85.46
83.85
24.42
80.42
75.60
94.68
87.75
88.40
59.55
84.09
82.00
87.50
82.23
82.40
98.36
85.09
85.90
92.93
83.66
84.15

198

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

Fig. 7. (a) The receiver operating characteristics (ROC) curves of the third subset of training and testing EEG signals by three proposed methods ( Q4WHF_SVD,, FIMF_EMD,
and Q4WHF_DCTHF_DWT,) against the EMD method using LS-SVM with RBF kernel from Frontal lobe and (b) Normalized distribution of feature values extracted from ictal
and interictal EEG signals using DCT, DWT, SVD, and IMF based method where 200 samples are taken from each of ictal and interictal categories in Frontal lobe.

best result compared to other IMFs. For the RBF kernel function,
the proposed methods outperform the state-of-the-art method
consistently and the average sensitivity using the proposed methods, such as Quartile four weighted high frequency component using
SVD (Q4WHF_SVD), Quartile four weighted high frequency component using DCT (Q4WHF_DCT), First IMF using EMD (FIMF_EMD),
and Quartile four weighted high frequency component using DCT and
high frequency from DWT (Q4WHF_DCTHF_DWT) are 83.33%,
99.12%, 96.47%, and 87.13% for RBF kernel, respectively, whereas,
the average sensitivity using the state-of-art-methods [17] and
[18] are 6.35% and 48.04%, respectively. Using Morlet kernel, the
average sensitivity of the proposed methods is higher compared to
the existing methods [17] and [18]. Moreover, we justify our result
applying linear kernel. Table 2 also presents that in most of the
cases the proposed methods with linear kernel provide better
results compared to the proposed methods with Morlet and RBF
kernels in terms of sensitivity. Table 2 also presents that the
proposed Q4WHF_DCT method is the best among all proposed
methods in terms of sensitivity. It is also interesting to note that
the accuracy of the proposed Q4WHF_DCTHF_DWT method is
reasonably consistence in terms of three criteria for RBF and
Morlet kernel. According to Table 2, all proposed methods always
outperform the existing methods [17,18] in terms of sensitivity (i.
e., seizure and ictal detection rate), however, in some cases the
existing methods [17,18] outperform one or more proposed
methods in terms of specicity (interictal detection rate). For
seizure detection, sensitivity is more important compared to the
specicity detection as the cost of failure detecting ictal is heavier
compared to interictal. In overall classication the proposed
methods are also better compared to the existing methods. We
combine all features to classify ictal and interictal EEG signals.
However, combining is not good because LS-SVM may not properly combine all available features to make a good prediction
model. The average sensitivity of the proposed methods is 91.36%
(average result of all kernel and lobe), whereas the sensitivities of
the state-of-the-arts methods are 54.61% [18] and 7.99% [17],
respectively.
The performance of the LS-SVM is evaluated by (ROC) plot
shown in Fig. 7(a), ROC illustrates the performance of a binary
classier system where it is created by plotting the fraction of true
positives from the positives, i.e. true positive rate (TPR) vs. the
fraction of false positives from negatives i.e., false positive rate
(FPR) with various threshold settings. TPR is known as sensitivity,
and FPR is one minus the specicity or true negative rate. Fig. 7(a)
shows that the proposed four methods are carrying good

classication results than that of the state-of-the-art method

[18] using dataset from Frontal lobe of the third subset of training
and testing EEG signals.
Fig. 7(b) shows the distribution of normalized feature values
(i.e., energy & entropy in the Q4WHF_DCT method, STD in the
Q4WHF_DCTHF_DWT method, STD & energy in the Q4WHF_SVD
method, and STD in the FIMF EMD method) of ictal and interictal
from the Frontal lobe dataset [22]. The gure shows that the range
of feature values for ictal is different from that of the interictal
signals. This phenomenon is the other evidence why we use those
features to classify ictal and interictal EEG signals in the proposed
methods.
First column of Fig. 8 shows classication comparison using the
proposed methods (i.e., FIMF EMD and Q4WHF_DCTHF_DWT)
and the state-of-the-art method [18]. Fig. 8(b) and (c) shows the
LS-SVM classication results using the proposed FIMF EMD and
Q4WHF_DCTHF_DWT against the state-of-the-art method (i.e.,
EMD Fig. 8(a)). We generate rst column of Fig. 8 from Frontal lobe
for the second subset of training and testing and obtain the
classication accuracy 91.20% for the Q4WHF_DCTHF_DWT
method, 87.20% for the FIMF_EMD method, and 80.80% for the
EMD method. It can be concluded from these gures that the
proposed method based on the DCTDWT and the IMF outperform
the state-of-the-art method (i.e., EMD). The same classication
comparisons are also provided in second column of Fig. 8 using the
state-of-the-art method based on the EMD and the proposed
method based on the SVD and DCT respectively. We consider
Temporal lobe for the fourth subset of training and testing; obtain
classication accuracy 91.20% for Q4WHF_SVD, 89.60% for
Q4WHF_DCT, and 85.20% for EMD [18].Thus, it is concluded that
the proposed methods based on DCT and SVD also outperform the
EMD method [18].

5. Conclusion
Unlike seizure and non-seizure EEG signals, classication of
ictal and interictal EEG signals is a challenging task due to non
abrupt phenomena (i.e., visually non-separable) between ictal and
interictal EEG signals. Moreover, the features of ictal and interictal
signals are not consistence in different brain locations, patient,
types of seizure, and hospital setting during epileptic period of a
patient. Thus, the classication performance for ictal and interictal
EEG signals using the existing features extraction and classication
techniques is not at expected level. Our initial observation

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

199

Fig. 8. First column represents the classication of ictal and interictal EEG signals from Frontal lobe for the second subset of training and testing using (a) the state-of-the-art
method against, (b) the proposed IMF and (c) the DCT-DWT method. Second column represents the classication of ictal and interictal EEG signals from Temporal lobe for the
fourth subset of training and testing using (d) the state-of-the-art method against (e) the proposed SVD and (f) the DCT methods.

indicates that high frequency components of EEG signals have

distinguished characteristics. To improve the accuracy, in this
paper we propose novel feature extractions strategies from
the high frequency components based on a number of established transformation and decompositions. The average sensitivity (which is the measure for seizure) of the proposed
methods is 91.36% (average result of all kernel and lobe)

whereas the sensitivity of the state-of-the-arts methods are

54.61% [18] and 7.99% [17], respectively. The experimental
results show that the proposed methods outperform the stateof-the-art methods in terms of accuracy, sensitivity, and specicity for the ictal and interictal EEG signals with greater
consistence from the large benchmark dataset in two different
brain locations.

200

M.Z. Parvez, M. Paul / Neurocomputing 145 (2014) 190200

References
[1] Neurology Applied: How Science is Bringing Music Instruction Back to
Expressive Development, URL: http://www.thomasjwestmusic.com/neurolo
gyapplied.htm, (accessed 27.08.12).
[2] EEG Signal, The science behind Bo-Tau, URL: http://www.bo-tau.co.uk/index.
php?action=the_science_behind_bo_tau, (accessed 27.08.12).
[3] Epilepsy-Seizures,URL:
http://www.healthcommunities.com/epilepsy-sei
zures/about-seizures.shtml, (accessed 27.08.12).
[4] M. Soleymani, T. Pun, M. Pantic, Multi-modal emotion recognition in response
to videos, IEEE Trans. Affect. Comput. 3 (2) (2012) 211223.
[5] S. Scholler, S. Bosse, M.S. Treder, B. Blankertz, G. Curio, K.-R. Mller,
T. Wiegand, Toward a direct measure of video quality perception using EEG,
IEEE Trans. Image Process. 21 (5) (2012) 26192629.
[6] W. Di, C. Zhihua, F. Ruifang, L. Guangyu, and L. Tian, Study on human brain
after consuming alcohol based on EEG signal, in: Proceedings of the International Conference on Computer Science and Information Technology (ICCSIT),
5, 406409, 2010.
[7] E. Estrada, H. Nazeran, F, Ebrahimi, and M. Mikaeili, EEG signal features for
computer-aided sleep stage detection, in: Proceedings of the IEEE EMBS
Conference on Neural Engineering, 669672, 2009.
[8] Z.M. Hanaah, K.F.M. Yunos, Z.H. Murat, M.N. Taib, S. Lias, EEG brainwave
pattern for smoking behaviour after horizontal rotation treatment, in: Proceedings of the IEEE Student Conference on Research and Development, 2009.
[9] Z.H. Murat, R. Shilawani, S.A. Kadir, R.M. Isa, and M.N. Taib, The effects of
mobile phone usage on human braingwave using EEG, in: Proceedings of the
13th International Conference on Modelling and Simulation, 2011.
[10] K. Lehnertz, R.G Andrzejak, J. Arnhold, T. Kreuz, F. Mormann, C. Rieke,
G. Widman, C.E. Elger, Nonliner EEG analysis in Epilepsy: its possible use for
interictal focus localization, seizure ancipation, and prevention, J. Clin.
Neurophsysiol. 18 (3) (2001) 209222.
[11] R. Panda, P.S. Khobragade, P.D. Jambhule, S.N. Jengthe, P.R. Pal, and T.K. Gandhi,
Classication of EEG signal using wavelet transform and support vector
machine for epileptic seizure diction, in: Proceedings of the International
Conference on Systems in Medicine and Biology (ICSMB), 405408, 2010.
[12] S.G. Dastidar, H. Adeli, N. Dadmehr, Mixed-band wavelet chaos-neural network methodology for epilepsy and epileptic seizure detection, IEEE Trans.
Biomeical Eng. 54 (9) (2007) 15451551.
[13] H. Ocak, Optimal classication of epileptic seizures in EEG using wavelet
analysis and genetic algorithm, Signal Process. 88 (2008) 18581867.
[14] K. Polat, S. Gunes, Classication of epileptiform EEG using a hybrid system
based on decision tree classier and fast Fourier transform, Appl. Math.
Comput. 187 (2) (2007) 10171026.
[15] S.F. Liang, H.C. Wang, W.L. Chang, Combination of EEG complexity and spectral
analysis for epilepsy diagnosis and seizure detection, EURASIP J. Adv. Signal
Process 853434 (2010) 115 (Article ID).
[16] R.B. Pachori, Discrimination between ictal and seizure-free EEG signals using
empirical mode decomposition, Res. Lett. Signal Process. 293056 (2008) 15
(Article ID).
[17] D. Birvinskas, V. Jusas, I. Martisius, and R. Damasevicius, EEG dataset reduction
and feature extraction using discrete cosine transform, in: Proceedings of the
European symposium on computer modeling and simulation (EMS), 199-204,
2012.
[18] V. Bajaj, R.B. Pachori, Classication of seizure and non-seizure EEG signals
using empirical mode decomposition, IEEE Trans. Inf. Technol. Biomed. 16 (6)
(2012) 11351142.
[19] Small dataset, Epilepsy dataset (text format), http://epileptologie-bonn.de/cms/
front_content.php?idcat=193&lang=3&changelang=3, (accessed 28.;04.12).
[20] J. Harrison, Fast and accurate bessel function computation, in: Proceedings of
the19th IEEE international symposium on computer arithmetic, 104113,
2009.
[21] X. Zhang, W. Diao, Z. Cheng, Wavelet transform and singular value decomposition of EEG signal for pattern recognition of complicated hand activities,
Lect. Notes Comput. Sci. Digit. Hum. Model. 4561 (2007) 294303.
[22] Large dataset, EEG Data set from Epilepsy Center of the University Hospital of
Freiburg http://epilepsy.uni-freiburg.de/freiburg-seizure-prediction-project/
eeg-database, (accessed 10.06.12).
[23] K.D. Brabanter, P. Karsmakers, F. Ojeda, C. Alzate, and J.D. Brabanter, K.
Pelckmans, B. De Moor, J. Vandewalle, J.A.K. Suykens, LS-SVMlab Toolbox
User's Guide, version 1.8, Katholieke Universiteit Leuven, 2011.
[24] S. Sanei, J.A. Chambers, EEG Signal Processing, John Wiley & Sons, Ltd., UK,
2007 (Centre of Digital Signal Processing, Cardiff University).
[25] S. Abe, Support Vector Machine for Pattern Classication, Second Edition,
Springer, London, Dordrecht, Heidelberg, New York, 2010.

[26] H. Wang and D. Hu, Comparison of SVM and LS-SVM for regression, in:
Proceedings of the international conference on neural networks and brain, 1,
279283, 2005.
[27] H. Xing, M. Ha, B. Hu, D. Tian, Linear feature-weighted support vector
machine, Fuzzy Inf. Eng. 1 (3) (2009) 289305.
[28] A Subasi, EEG signal classication using wavelet feature extraction and a
mixture of expert model, Expert Syst. Appl. 32 (4) (2007) 10841093
(Elsevier).
[29] J.R. Williamson, D.W. Bliss, D.W. Browne, J.T. Narayanan, Seizure prediction
using EEG spatiotemporal correlation structure, Epilepsy Behav. 25 (2) (2012)
230238.
[30] TP. Jung, S. Makeig, C. Humphries, et al., Removing electroencephalographics
artifacts by blind source separation, Psychophysiology 37 (2000) 163178.
[31] W. De Clercq, A. Vergult, B. Vanrumste, W.V. Paesschen, S.V. Huffel, Canonical
correlation analysis applied to remove muscle artifacts from the electroencephalogram, IEEE Trans. Biomeical Eng. 53 (12) (2006) 25832587.
[32] J. Ma, P. Tao, S. Bayram, V. Svetnik, Muscle artifacts in multichannel EEG:
characteristics and reduction, Clin.Neurophysiol. 123 (8) (2012) 16761686.
[33] G.A. Worrell, L. Parish, S.D. Cranstoun, R. Jonas, G. Baltuch, B. Litt, Highfrequency oscillations and seizure generation in neocortical epilepsy, Brain
127 (2004) 14961506.
[34] M. Paul, W. Lin, C.T. Lau, B.S. Lee, A long term reference frame for hierarchical
b-picture based video coding, IEEE Trans. Circuits Syst. Video Technol. (2014),
http://dx.doi.org/10.1109/TCSVT.2014.2302555.
[35] Z. Gu, W. Lin, C.T. Lau, B.S. Lee, and M. Paul, Two dimensional singular value
decomposition (2D-SVD) based video coding, in: Proceedings of the IEEE
conference on image processing (IEEE ICIP-10), 181184, 2010.

Mohammad Zavid Parvez received B.Sc.Eng. (hons.)

Degree in Computer Science and Engineering from
Asian University of Bangladesh in 2003 and M.Sc.
Degree in Electrical Engineering with emphasis on
Signal Processing from Blekinge Institute of Technology,
Sweden in 2010. Parvez's research interests include
biomedical signal processing and machine learning.
He has more than three year industry experience as
a software developer. Moreover, he was involved several software projects in Denmark. Currently he is a
PhD student and teaching staff in Charles Sturt University.
Parvez is a Graduate student member of IEEE. He has
received full funded scholarship from the Faculty. He has published several journal
and conference papers in the area of EEG signal analysis and cognitive radio.

Manoranjan Paul received B.Sc.Eng. (hons.) Degree in

Computer Science and Engineering from Bangladesh
University of Engineering and Technology (BUET), Bangladesh, in 1997 and PhD Degree from Monash University, Australia in 2005. He was an Assistant Professor
in Ahsanullah University of Science and Technology. He
was a Post-Doctoral Research Fellow in the University
of New South Wales, in 20052006, Monash University,
in 20062009, and Nanyang Technological University,
in 20092011. He has joined in the School of Computing and Mathematics, Charles Sturt University (CSU) at
2011. Currently he is a Senior Lecturer and Associate
Director of the Center for Research in Complex Systems
(CRiCS) in CSU.
Dr Paul is a senior member of the IEEE and ACS. He is in editorial board of three
international journals including EURASIP Journal on Advances in Signal Processing
(JASP). Dr. Paul has served as a guest editor of Journal of Multimedia and Journal of
Computers for ve special issues. He has been in Technical Program Committees of
more than 25 international conferences. Dr Paul obtained more than A$1M
competitive grant money including Australian Research Council (ARC) Discovery
Project. He has published more than 65 refereed publications including journals
and conferences. He organized a special session on New video coding technologies in IEEE ISCAS 2010. He was a keynote speaker on Vision friendly video
coding in IEEE ICCIT 2010 and tutorial speaker on Multiview Video Coding in
DICTA 2013. His major research interests are in the elds of video coding, computer
vision, and EEG Signals analysis. Dr Paul received Vice Chancellor Research
Excellence Award in Faculty level at 2013.