Computational Intelligence - February 2014 - 53

Ching-Kun Chen and Chun-Liang Lin
Department of Electrical Engineering,
National Chung Hsing University,
Taichung 402, TaiwaN
Shyan-Lung Lin
Department of automatic Control
Engineering, Feng Chia University,
Taichung, 407, TaiwaN
Yen-Ming Chiu
Department of Electrical Engineering,
National Chung Hsing University,
Taichung 402, TaiwaN
Cheng-Tang Chiang
Professional Medical Corporation,
BSN Group, Taichung 402, TaiwaN
I. Introduction

ophisticated technologies realized
from applying the idea of biometric identification are increasingly
applied in the entrance security management system, private document protection, and security access control.
Common biometr ic identification
involves voice, attitude, keystroke, signature, iris, face, palm or finger prints,
etc. Still, there are novel identification
technologies based on the individual's
biometric features under development
[1-4].
ECGs are commonly adopted in
medical diagnosis. The voltage level,
waveform and interval time in ECG signals are important features for medical
diagnosis. However, the explanation of
the complicated ECGs for the diagnosis
purpose highly relies on doctors' expertise and experience. Therefore, quantifying the ECG signals in depth is a must
for the automated diagnosis. Many studies have shown that the features of ECG
signals of normal people are different
from the people with heart diseases
[5-20]. In time-domain, Tsipouras et al.
[5] adopted the R-R interval signal of
the ECG and proposed a simple method
to do the analysis. In frequency-domain,
Chazal et al. [6] used a wavelet transform to classify the electrocardiogram.
The heart rate variability of ECG has
been studied and analyzed in [7-9].
Some researchers repeated the past
researches by utilizing the evidence of

Digital Object Identifier 10.1109/MCI.2013.2291691
Date of publication: 14 January 2014

A Chaotic Theoretical Approach
to ECG-Based Identity Recognition
difficult to duplicate. Therefore, an
chaos characteristics and showed that a
ECG signal can be a biometric tool for
human heart is not a simple oscillator.
individual identification [22-33].
Actually the heart behavior exhibits a
Recently, the ECG has been
chaotic behavior [10]. To analyze the
attempted in the field of biomedical
reconstructed phase plot, Lyapunov
engineering for personal
parameters and cor relation
identity verification,
dimension parameter s
see Table 1. In prewere shown to be usev i o u s wo r k s ,
ful to understand the
some of tested
complex heart
data were taken
behavior. In the litfrom the availerature, Babloyants et
able database such
al. [10] analyzed the
as MIT-BIH/PTB,
ECG of four healthy
and some were actually
people in detail. Fojt et al.
© brand x pictures
collected in the laboratory. The rec[11] calculated the correlation dimenognition rate is only for reference
sion parameters of seven healthy people.
because all approaches were not conLater, the parameters of the Lyapunov
ducted under the same conditions such
exponents and correlation dimension
as the tested subjects, data source, elecwere utilized to identify different physitrode orientation etc.
ology conditions [12] or heart disease
Basically, the approaches can be
[13-15]. In some researches, classificadivided into two major types concerntions of the heart disease by the reconing feature extraction from ECG, i.e.
struction phase plot have also been
fiducial detection [22-25] and non-fiduattempted [16-20].
cial detection [26-33]. The former relies
Biometric based authentication syson time duration in the single heartbeat
tem provides better security solutions
waveform. There is a weakness which
than before, but some biological
the detection points need to be precise
parameters that are used as biometric
since a slightest variation of the fiducial
do not provide the robustness against
point location may result in misclassififalsified credentials such as voice which
cations. The misclassification maybe
can be copied through microphone,
even more severe since every device
fingerprints which can be collected on
manufacturer follows its own definition
silicon surfaces and iris which can be
of ECG wavelength boundaries that
copied on contact lenses. ECG does
limits the scope of applications. On the
not have these problems and it is
contrary, the non-fiducial approaches
unique in every individual [21]. Compurpose to extract discriminative inforpared with the common biometricmation from ECG signals without havbased systems, ECG signals change
ing to localize the fiducial points.
from person to person and is extremely

February 2014 | ieee Computational intelligenCe magazine

Table of Contents for the Digital Edition of Computational Intelligence - February 2014