American Journal of Engineering Research (AJER) 2013

American Journal of Engineering Research (AJER)

e-ISSN: 2320-0847 p-ISSN : 2320-0936
Volume-02, Issue-03, pp-44-50
www.ajer.us

Research Paper Open Access

Dwt - Based Feature Extraction from ecg Signal

V.K.Srivastava1, Dr. Devendra Prasad2
1
Research Scholar, Deptt. of Electronics & Comm. Engg., JJT University, Jhunjhunu, Rajasthan, India
2
Associate Professor, Deptt. of Computer Sc.& Engg., M.M. University, Ambala, Haryana,India

Abstract: Electrocardiogram is used to measure the rate and regularity of heartbeats to detect any irregularity to the heart.
An ECG translates the heart electrical activity into wave-line on paper or screen. For the feature extraction and classification
task we will be using discrete wavelet transform (DWT) as wavelet transform is a two-dimensional timescale processing
method, so it is suitable for the non-stationary ECG signals(due to adequate scale values and shifting in time). Then the data
will be analyzed and classified using neuro-fuzzy which is a hybrid of artificial neural networks and fuzzy logic.

Keyword: Electrocardiogram (ECG), DWT, Neuro Fuzzy.

I. INTRODUCTION
To attain a progressive sustainable development in all regions of the world, attention should be diverted to
the health of a population. All sciences contribute to the maintenance of human health and the practice of medicine.
Medical physicists and biomedical engineers are the professionals who develop and support the effective utilization of
this medical science and technology as their responsibilities to enhance human health care with the new development
of the medical tools such as electrocardiogram (ECG). Heart disease is a broad term that includes several more
specific heart conditions which are Coronary Heart Disease, Heart Attack, Acute Coronary Syndrome, Aortic
Aneurysm and Dissection, Ischemia, Arrhythmias, Cardiomyopathy, Congenital Heart Disease, Peripheral Arterial
Disease (PAD).The most common heart condition is coronary heart disease, which can lead to heart attack and
other serious conditions [22] and the Myocardial Ischemia is the most common cause of death in the
industrialized countries [9].
The electrocardiogram (ECG) is a noninvasive and the record of variation of the biopotential signal of
the human heartbeats. The noninvasive technique meaning that this signal can be measured without entering the
body at all. Electrodes are placed on the users skin to detect the bioelectric potentials given off by the heart that
reach the skins surface. The ECG detection which shows the information of the heart and cardiovascular
condition is essential to enhance the patient living quality and appropriate treatment. It is valuable and an
important tool in the diagnosing the condition of the heart diseases.
In recent year, numerous research and algorithm have been developed for the work of analyzing and
classifying the ECG signal. The classifying method which have been proposed during the last decade and under
evaluation includes digital signal analysis, Fuzzy Logic methods, Artificial Neural Network, Hidden Markov
Model, Genetic Algorithm, Support Vector Machines, Self-Organizing Map, Bayesian and other method with
each approach exhibiting its own advantages and disadvantages.
The ECG features can be extracted in time domain [5][4] or in frequency domain [8][9]. Some of the
features extraction methods implemented in previous research includes Discrete Wavelet Transform [4],
Karhunen-Loeve Transform [11], Hermitian Basis and other methods [10].

www.ajer.us Page 44
 American Journal of Engineering Research (AJER) 2013

Figure 1: An example of normal ECG trace

All of the ECG waveforms pattern and variability must be determine accurately to get the better
diagnostic result that will shown the correct heart disease of the patient. Figure 1 shows the normal ECG traces
which consist of P wave, QRS complex and T wave. The P wave is the electrical signature of the current that
causes atrial contraction, the QRS complex corresponds to the current that causes contraction of the left and
right ventricles and the T wave represents the repolarization of the ventricles.
This research is mainly aimed to extract important parameters from the ECG signal through the DWT
technique. By applying this signal analysis technique, the most important parameter of the ECG signal can be
taken as the analysis data. Then, the data will be used as an input to the classifier to identify the heart disease.
The idea of the ECG analysis and classification using Neuro Fuzzy has been start around 1990, yet it remains
one of the most important indicators of proper heart disease classification today.
The most difficult problem faced by an automatic ECG analysis is the large variation in the morphologies of
ECG waveforms, it happen not only for different patients or patient groups but also within the same patient. So the
Neuro Fuzzy is the most suitable technique because it is more tolerance to morphological variations of the ECG
waveforms. This project is useful for the medical application enhancement such as in hospital, clinic which can
automatically help to increase the patient health care.

II. . RELATED RESEARCH

The Wavelet transform is a two-dimensional timescale processing method for non-stationary signals
with adequate scale values and shifting in time [17]. The wavelet transform is capable of representing signals in
different resolutions by dilating and compressing its basis functions [18]. The researcher Cuiwei Li et al [4] use
the Wavelet transforms method [19][20][21] because the results indicated that the DWT-based feature extraction
technique yields superior performance. Neuro-fuzzy is a hybrid of artificial neural networks and fuzzy logic [1].
Fuzzy Neural Network as in the literature [1] incorporates the human-like reasoning style of fuzzy systems
through the use of fuzzy sets and a linguistic model consisting of a set of IF-THEN fuzzy rules. The learning
process of FNN consists of three phases which are fuzzy membership, fuzzy rule identification and supervised
fine-tuning.
Neuro-Fuzzy approach consist of a five-layer neural network called input linguistic layer, condition
layer, rule layer, consequent layer, output linguistic layer. The fuzzification of the inputs and the defuzzification
of the outputs are performed in the input linguistic stage and output linguistic layers performed by the fuzzy rule
identification.
The performance enhancement using proposed method in Neuro-Fuzzy using autoregressive model
coefficients, higher-order cumulant and wavelet transform variances as features had been used by Mehmet
Engin [15]. It used self-organizing layer which responsible for the clusterization of the input data which is
feature vector and the outputs of the membership values form the input vector to the second sub-network
(MLP).
This research use the output of DWT technique as features vector and Neuro-Fuzzy as the classifier for
the ECG analysis, because based on the previous research, the accuracy rates achieved by the combined neural
network model presented for classification of the ECG beats were to be higher than stand alone classifier model.

III. PROPOSED SYSTEM

www.ajer.us Page 45
 American Journal of Engineering Research (AJER) 2013
Various methodologies of automated diagnosis have been done. The ECG data analysis techniques are
reviewed in and evaluate proposed methods of the classification methods. In our proposed, the process of ECG
analysis can be generally subdivided into a number of disjoint processing modules have been identified which
are:

1. Data acquisition
2. Pre-processing beat detection
3. Feature Extraction
4. Classification

A. Data acquisition
1. Data Collection
The heart sound using ECG will record simultaneously from patients. The ECG Database in PhysioBank also
being used as a data of ECG input signal for features extraction part processing.

2. Preparation of Data
Partitioning the ECG signal into cardiac cycles, and detection of the main events and intervals in each
cycle. Major features such as the QRS amplitude, R-R intervals, waves slope of ECG signal can be used as
features to create the mapping structure.

B. Preprocessing
The first step of signal pre processing is filtering the ECG signal because as any other measured signal,
ECG is also contaminated with high frequency noise. The unwanted noise of the heart biopotential signal must
be removing. ECG were filtered using a bandpass filter between 0.05HZ-100Hz to eliminate the motion artifact,
baseline wander and 50Hz notch filter to eliminate power line noise.

C. Features Extraction using DWT

The Wavelet Transform (WT) is designed to address the problem of non-stationary ECG signals. It derived
from a single generating function called the mother wavelet by translation and dilation operations. The main
advantage of the WT is that it has a varying window size, being broad at low frequencies and narrow at high
frequencies, thus leading to an optimal time-frequency resolution in all frequency ranges.
The WT of a signal is the decomposition of the signal over a set of functions obtained after dilation and
translation of an analyzing wavelet [13]. The ECG signals which consisting of many data points, can be
compressed into a few features by performing spectral analysis of the signals with the WT. These features
characterize the behavior of the ECG signals. Using a smaller number of features to represent the ECG signals is
particularly important for recognition and diagnostic purposes [14].
The ECG signals were decomposed into time-frequency representations using Discrete Wavelet Transform
(DWT). The DWT technique has been widely used in signal processing tasks in recent years. The major advantage of
the DWT is that it provides good time resolution. Good resolution at high frequency and good frequency resolution at
low frequency. Because of its great time and frequency localization ability, the DWT can reveal the local
characteristics of the input signal.
The DWT represents a 1-Decompodition signal s(t) in terms of shifted versions of a low pass scaling
function φ(t) and shifted and dilated versions of a prototype bandpass wavelet function ψ(t).

(-j/2) -j
Ψj,k(t) = 2 ψ (2 t –k) (1)
-j -j
φ j,k (t) = 2 φ (2 t –k) (2)

where: j controls the dilation or translation

k denotes the position of the wavelet function

Discrete Wavelet Transform is also referred to as decomposition by wavelet filter banks. This is
because DWT uses two filters, a low pass filter (LPF) and a high pass filter (HPF) to decompose the signal into
different scales. The output coefficients of the LPF are called approximations while the output coefficients of
the HPF are called details. The approximations of the signal are what define its identity while the details only
imparts nuance.

www.ajer.us Page 46
 American Journal of Engineering Research (AJER) 2013
The DWT can be calculated at different resolutions using Mallat-algorithm to utilize successive lowpass and
highpass filters to compute DWT.

The procedure of DWT decomposition of a input signal x[n] is schematically shown in the Figure 3 above.
Each stage consists of two digital filters and two downsamplers by 2 to produce the digitized signal. The first filter,
g[n] is the discrete mother wavelet, which is high-pass filter, and the second, h[n] is low-pass filter. The downsampled
outputs of first high pass filters and low-pass filters provide the detail, D1 and the approximation, A. The first
approximation, A1 is decomposed again and this process is continued. The decomposition of the signal into
different frequency bands is simply obtained by successive highpass and lowpass filtering of the time domain
signal. The signal decomposition can mathematically be expressed as follows:

yhi [k] = Σ x[n].g[2k – n] (3)

ylo[k] = Σ x[n].h[2k – n] (4)

D. ECG Classification using Neuro Fuzzy

In the classifier stage, the idea to apply fuzzy hybrid neural network which composed of two
subnetworks connected in cascade. The cascading of the fuzzy self-organizing layer performing the pre
classification task and the Multilayer Perceptron (MLP) of Neural Network working as the final classifier which
is the fuzzy rules are modeled by artificial neural networks (ANNs)
In the fuzzy self-organizing layer, it is responsible to analyze the distribution of data and group the data into
the different membership values. This membership value is applied as the input vector to the MLP Neural Network
classifier. The membership value also representing the parameter of each heart beat class. The used of Fuzzy neural
network has a number of advantages when compared to ordinary learning of neural networks class. The fuzzy neural
network is more tolerant to the noise and less sensitive to the morphological changes of the ECG characteristics.

www.ajer.us Page 47
 American Journal of Engineering Research (AJER) 2013
E. Neuro Fuzzy Model

A Neuro Fuzzy model contains three main components, which are fuzzification stage, the rule base and the
defuzzification stage.

1. The first layer represents the direct input layer. Each node represents one ECG feature and sends this input variables
from the features extraction to next layer directly.
2. The second layer is the fuzzification layer. In this layer, each extracted ECG feature is fed to fuzzy nodes.
These nodes perform to transform the crisp values of the input variables into fuzzy membership values. It
represent the membership degree of the input ECG feature to its linguistic description.
3. The third layer represents the fuzzy rules stage. Each node in this layer represents one fuzzy if-then rule.
4. The last layer is the neural network connecting layer. The

linguistic description will shows the ECG classification result of each lead group. Based on the Mamdani model
[16], the fuzzy system by existing knowledge described as:

i i i i
Rule i, R : IF x 1 is A 1 and x2 is A 2 …and x n is A n,
i i
THEN y is C (5)

i
Where R (I =1,2…,n) denotes the ith fuzzy rules. x=(x1,x2,.. .,xn) is the input vector,
i i
y is the output of the fuzzy rule R
i
A j (j=1,2, ..., n) are fuzzy membership function (linguistic description)

The performance measures used to evaluate the performance of the classifiers were accuracy, sensitivity, specificity.
(i) Specificity = number of correct classified normal Number of total normal beats
(ii) Sensitivity =number of correct classified heart disease Number of total heart disease
(iii) Accuracy = Number of beats correctly classified Total number of beats tested

www.ajer.us Page 48
 American Journal of Engineering Research (AJER) 2013
IV. RESULT
The experiment result of ECG analysis using MATLAB programming has been done to show the
preprocessing of biosignal analysis before the features extraction process. The ECG Database from Holy
Hospital has been used for ECG input signal of the preprocessing stage.

Figure 6: Original ECG signal Figure 7: Baseline Elimination of ECG signal

Figure 8: Flow of ECG signal process

V. CONCLUSION
This paper has presented the system which based on the application of Neuro Fuzzy network for
reliable heartbeat classification on the basis of the ECG waveform. The DWT characterization will delivers the
stable features to the morphology variations of the ECG waveforms. The potential of accuracy rate for Neuro
Fuzzy system as ECG diagnostic decision aid is very high.

www.ajer.us Page 49
 American Journal of Engineering Research (AJER) 2013
REFERENCES
[1]. Yuksel Ozbaya,Rahime Ceylan, Bekir Karlik, “A Fuzzy Clustering Neural Network Architecture For classification
Of ECG Arrhythmias”, Department of Electrical & Electronics Engineering, Selcuk University, Konya, Turkey, Computers in
Biology and Medicine 376–388, 2006.
[2]. Stanislaw Osowski and Tran Hoai Linh, “ECG Beat Recognition Using Fuzzy Hybrid Neural Network”, IEEE Transactions on
Biomedical Engineering, Vol. 48, No. 11, November 2001.
[3]. Tran Hoai Linh, Stanisaw Osowski and Maciej Stodolski, “On-Line Heart Beat Recognition Using Hermite Polynomials and Neuro-
Fuzzy Network”, Warsaw University of Technology, 2003.
[4]. Cuiwei Li, Chongxun Zheng, Changfeng Tai, “Detection of ECGCharacteristic Points Using Wavelet Transforms”, IEEE Transactions
on Biomedical Engineering, Vol. 42, No. 1, January 1995.Hu, Y.H., Palreddy, S., Tompkins, W., “A Patient Adaptable ECG Beat
Classifier Using A Mixture Of Experts Approach”, IEEE Trans.Biomed. Eng. (44), 891–900, 1997.
[5]. De Chazal, P., Reilly, R.B., “Automatic Classification of ECG Beats Using Waveform Shape And Heart Beat Interval Features” , In
Proceedings of the IEEE International Conference on Acoustic, Speech and Signal Processing (ICASSP’03), Hong Kong, China,
2003.
[6]. Minami, K., Nakajima, H., Toyoshima, T., “Real-Time Discrimination of Ventricular Tachyarrhythmia with Fourier-Transform
Neural Network”, IEEE Trans. Biomed. Eng. (46), 1999
[7]. Moraes, J.C.T.B., Seixas, M.O., Vilani, F.N., Costa, E.V., “A Real Time QRS Complex Classification Method Using Mahalanobis
Distance” Comput. Cardiol., 2002.
[8]. De Chazal, P., Reilly, R.B., “Automatic Classification of ECG Beats Using Waveform Shape And Heart Beat Interval Features” , In
Proceedings of the IEEE International Conference on Acoustic, Speech and Signal Processing (ICASSP’03), Hong Kong, China,
2003.
[9]. Minami, K., Nakajima, H., Toyoshima, T., “Real-Time Discrimination of Ventricular Tachyarrhythmia with Fourier-Transform
Neural Network”, IEEE Trans. Biomed. Eng. (46), 1999
[10]. Moraes, J.C.T.B., Seixas, M.O., Vilani, F.N., Costa, E.V., “A Real Time QRS Complex Classification Method Using Mahalanobis
Distance” Comput. Cardiol., 2002.
[11]. Costas Papaloukas, Dimitrios I. Fotiadis, Aristidis Likas, Lampros K. Michalis, “Automated Methods for Ischemia Detection in
Long Duration ECGs”, 2003.
[12]. David Cuesta-Frau, Juan C, Perez-Cortes, Gabriela Andrea Garcia, Daniel Navak, “ Feature Extraction Methods Applied to the
Clustering of Electrocardiographic Signals. A Comparative Stucy”, Czech Technical University In IEEE Proceeding, 2002. Franc
Jager, “Feature Extraction and Shape Representation of Ambulatory Electrocardiogram Using the Karhunen-Lo`eve Transform”
Electrotechnical Review, Ljubljana, Slovenija, 2002.
[13]. Inan Gulera,Elif Derya Ubey, “Ecgbeat Classifier Designed By Combined Neural Network Model”, Pattern Recognition, 199 – 208,
2005.
[14]. I. Daubechies, “The Wavelet Transform, Time-Frequency Localization and Signal Analysis, IEEE Trans. Inform. Theory, 961–
1005, 1990. Z. Dokur and T. Olmez, “ECG Beat Classification By A Novel Hybrid Neural Network, Comput. Meth. Prog. Biomed,
167–181, 2001.
[15]. Mehmet Engin, “Ecg Beat Classification Using Neuro-Fuzzy Network”, Pattern Recognition Letters 25, 1715–1722, 2004.
[16]. E. H. Mamdani and S. Assilian, “An Experiment In Linguistic Synthesis With A Fuzzy Logic Controller,” Int.J. Man-Mach Stud.,
vol. 7, pp. 1–13, 1975.
[17]. Thakor, N.V, “Multiresolution Wavelet Analysis Of Evoked Potentials”, In IEEE Trans. Biomedical. Eng. 40 (11), 1993.
[18]. Clarek, I, “ Multiresolution Decomposition Of Non-Stationary EEG Signal; A preliminary study”, Computer Bio.Medical. 25 (4),
1995.
[19]. S. Z. Mahmoodabadi, A. Ahmadian, D. Abolhasani, M. Eslami, J. H. Bidgoli, “ECG Feature Extraction Based on Multiresolution
Wavelet Transform”, In Proceedings of the 2005 IEEE Engineering in Medicine and Biology 27th Annual Conference Shanghai,
China, September 1-4, 2005.
[20]. Louis C Pretorius and Cobus Nel, “Feature Extraction From ECG for Classification By Artificial Neural Network”, University of
Pretoria, In IEEE Proceeding, 2002.
[21]. Ming-Yao Yang, Wei-Chih Hu and Liang-Yu Shyu, “ECG Events Detection and Classification Using Wavelet and Neural
Networks”, In IEEE Proceedings 19th International Conference, Oct to Nov, 1997.
[22]. Joe-Air Jiang, Chih-Feng Chao, Ming-Jang Chiu, Ren- Guey Leec,Chwan-Lu Tsengd, Robert Lin,“An Automatic Analysis Method
For Detecting And Eliminating ECG Artifacts In EEG”, Computers in Biology and Medicine, January 2006.

www.ajer.us Page 50