Classification of Normal and Epileptic Seizure EEG Signals Based on Empirical Mode Decomposition - Complex System Modelling and Control Through Intelligent Soft Computations - page 363

epilepsy patients can be useful for diagnosis of epilepsy (Andrzejak et al. 2001 ).
The presence of interference or artifacts due to external sources, in the EEG signal
recording may create problem in the diagnosis based on these recorded EEG sig-
nals. Therefore,

filtering is required to remove these artifacts (Senthil et al. 2008 ).

1.1 Epileptic Seizure EEG Signals

Epilepsy is one of the most common neurological disorders of human brain. As
epileptic activity manifests the clear and abnormal transient patterns in a normal
EEG signal, therefore EEG signals are widely used in diagnostic application for
detection of epilepsy. In epileptic patients, brain exhibits the process known as
'

(Cross and Cavazos 2007 ) in which normal neural network
abruptly converts into a hyper-excitable network, causing evocation of strange
sensations and emotions or sometimes muscle spasms and consciousness loss. In
such subjects, the nerve cells in the brain transmit excessive electrical impulses that
cause epileptic seizures. Epilepsy is recognized by occurrence of such unprovoked
seizures. Evaluation of the epilepsy can be performed by recording and analyzing
the epileptic seizure EEG signals from the electrodes which are placed on the
affected area on the brain scalp region (Coyle et al. 2010 ; Ince et al. 2009 ). The
recorded EEG signals are complex, non-linear, and non-stationary in nature
(Acharya et al. 2013 ; Boashash et al. 2003 ; Pachori and Sircar 2008 ; Pachori 2008 ).
The epileptic seizures can have severe harmful effect on the brain. Manual process
to identify the seizure events, consists of visual inspection and review of the entire
recorded EEG signals by trained expert, which is time consuming process and
demands considerable skills. Moreover, subjective nature of expert can also affect
the judgement of seizure events in EEG records. Therefore, it is appealing to
develop computer-aided automatic analysis method that consists of advanced signal
processing techniques, for classi

epileptogenesis '

cation between normal and epileptic seizures EEG

signals in recorded EEG signals.

1.2 Classi cation of Epileptic Seizure EEG Signals

Various methods have been developed for automatic classi cation of the epileptic

seizures by extracting parameters from EEG signals. These parameters can be
extracted using time-domain, frequency-domain, time-frequency domain and non-
linear methods of analysis and serve as the features for classi cation of EEG signals

based on signal processing methods (Acharya et al. 2013 ).
Many time-domain based techniques have been presented in literature with an
objective to detect epileptic seizures from EEG signals. The value of linear pre-
diction error energy has been found to be much higher in seizure EEG signals than