Improved Human Identification Method Based on
Electrocardiogram using Ensemble Empirical Mode
Decomposition and Teager Energy Operator
Yanjun Deng
1
, Zhidong Zhao
2
, Yefei Zhang
1
, Diandian Chen
3
1.
Sc
hool of Communication Engineering
Hangzhou Dianzi University
Hangzhou 310018, China
2. Hangdian Smart city research center of Zhejiang
Province
Hangzhou Dianzi University
Hangzhou 310018, China
3. College of Information Engineering
China Jiliang University
Hangzhou 310018, China
Abstract—The purpose of this research is to develop a biometric
system for individual identification with the electrocardiogram
(ECG) signal. The ECG signal varies from person to person and
it can be used as a new biometric for individual identification.
This paper presents a robust preprocessing stage to eliminate
the effect from noise and heart rate. A new feature extraction
technique known as Ensemble Empirical Mode Decomposition
(EEMD) with Teager Energy Operator (TEO) is derived and
used to generate novel ECG feature vectors. The dimensionality
reduction method Principal Component Analysis (PCA) is used
reduce the feature space before classification. Finally, K-Nearest
Neighbor (KNN) and Support Vector Machine (SVM) algorithm
are chosen as the classifiers. The proposed method is validated
by experiments on 40 subjects from three public databases; the
experiment results show that the subject recognition rate
achieves 95.5% and 97.5% with KNN and SVM classifier
respectively. For larger changes in heart rate, it also shows
strong stability.
Keywords-human identification; EEMD; TEO;
I.
I
NTRODUCTION
Automatic and accurate identity validation is becoming
increasingly critical in several aspects of our daily lives such
as in financial transactions, access control and others.
Biometric recognition was introduced as a more secure means
of identity establishment. Biometric features are unique for
every individual and that can be used to establish his/her
identity in a population. Among all of the biometric features,
the advantages of using ECG for biometric recognition can be
summarized as universality, permanence, uniqueness,
robustness to attacks, liveness detection, continuous
authentication and data minimization.
Prior works in the ECG biometric recognition field can be
categorized as fiducial point dependent or independent.
Fiducial-point approaches depend on local characteristics of
the heartbeat. The non-fiducial point approaches extract
features statistically, based on the overall morphology of the
waveform.
In this paper, we propose a new feature extraction
approach based on EEMD and TEO for automatic analysis on
single lead ECG for application in biometric identification.
The proposed system utilizes a robust preprocessing stage
applying on the raw ECG signal to handle the noise and HRV.
The dimensionality reduction method PCA is used reduce the
feature space before classification. Finally, the extracted
features are sending to the classifiers of KNN and SVM.
II. M
ETHODS
A
ND
M
ATERIALS
A. Electrocardiogram data
ECG signals collected in clinical situations include noise
resulting from such as baseline drift, power line interference,
and electromyography (EMG) interference. These noise alters
the waveform of the ECG trace from its original
structure, greatly reducing the ECG signal-to-noise ratio and
decreasing the reliability of ECG waveform detection.
Typically, the heart rate at a normal sinus rhythm is 60–
100 beats per minute (bpm). During ECG recording, the signal
is significantly affected by factors such as exercise, shock,
body chemistry, or emotions such as stress and anxiety, which
alter the morphology of the ECG because of fluctuations in
the heart rate. These changes in the external environment are
directly reflected in the ECG signal waveform. However, the
PQ segment and QRS wave are stable despite the change in
heart rate, whereas the T wave varies with the change in heart
rate.
B. Identification method
The proposed identification approach consists of three
main stages: preprocessing, feature extraction, and
classification. A block diagram of the identification system is
shown in FIGURE 1.
In the proposed identification approach, the noise is
eliminated from the original ECG and the heartbeat is
normalized to eliminate the HRV. The signal is then
decomposed using the ensemble empirical mode
decomposition (EEMD) method, and Teager energy operator
(TEO) is used to generate feature vectors. Because the
dimensionality of feature space is considerably high and
inappropriate for an efficient system, the PCA is used to
reduce the dimensionality. Classification of the reduced
feature coefficients is performed by the SVM and k-nearest
neighbor (K-NN) classifiers using Euclidean distance.
2017 10th International Congress on Image and Signal Processing, BioMedical Engineering and Informatics( CISP-BMEI 2017)
978-1-5386-1936-0/17/$31 ©2017 IEEE