IAES International Journal of Artificial Intelligence (IJ-AI)

Vol. 10, No. 4, December 2021, pp. 960~970

ISSN: 2252-8938, DOI: 10.11591/ijai.v10.i4.pp960-970  960

Electrocardiogram signals classification using discrete wavelet

transform and support vector machine classifier

Youssef Toulni1, Taoufiq Belhoussine Drissi2, Benayad Nsiri3

1,2Laboratory Industrial Engineering, Information Processing, and Logistics (GITIL), Faculty of Science Ain Chock,
University Hassan II-Casablanca, Morocco
1,3Research Center STIS, M2CS, National School of Arts and Crafts of Rabat (ENSAM), Mohammed V University in

Rabat, Morocco

Article Info ABSTRACT

Article history: The electrocardiography allowed us to make a diagnosis of several
cardiovascular diseases by representing the electrical activity of the heart over
Received Nov 29, 2020 time; this representation is called the electrocardiogram (ECG) signal. In this
Revised Sep 10, 2021 study we have proposed a model based on the processing of the ECG signal
Accepted Sep 28, 2021 by the wavelet decomposition using discrete wavelet transform (DWT). This
decomposition firstly makes it possible to denoise the signal then to extract
the statistical features from the approximation coefficients of the denoised
Keywords: signal and finally to classify the data obtained in a support vector machine
(SVM) classifier with cross validation for more credibility. After having tested
Cardiovascular disease this model with different mother wavelets at different scales, the accuracies at
Discrete wavelet transform the fourth scale are high and the best accuracy obtained is 87.50%.
Electrocardiogram signal
Support vector machine
This is an open access article under the CC BY-SA license.

Corresponding Author:
Youssef Toulni
Laboratory Industrial Engineering, Information Processing, and Logistics (GITIL)
Faculty of Science Ain Chock
Km 8 Route d'El Jadida, B.P 5366 Maarif Casablanca 20100 Morocco
Email: [email protected]

1. INTRODUCTION
Cardiovascular disease is a collection of irregularities affecting the heart; it is considered one of the
most important causes of death in the world. According to the World Health Organization, there were
approximately 17.9 million deaths in 2016; this big number represents 31% of deaths worldwide [1]. The
lives of people with cardiovascular diseases are in constant danger, quick and effective diagnosis of these
diseases can save a lot of lives. Several techniques in the medical field are used to diagnose cardiovascular
disease, such as blood tests, coronary angiography, cardiac MRI, X-ray and electrocardiography. However,
most of these techniques require medical assistance from experienced people, which is not always the case if
we knew that almost 30% of cases with these diseases come from poor countries. Electrocardiography is a
non-invasive detection technique based on recording the electrical activity of the heart over time [2], the
signal obtained during this recording is called an electrocardiogram (ECG). The ECG signal is considered
among the most widely used biomedical signals to detect heart problems, ECG signal contains a large
number of information that can be of great interest in the detection and diagnosis of many heart diseases [3]
which appears in some distortions of the signal shape. Despite the advantages of the use of ECG signals,
there are many limitations of this technique; the difficulty of interpreting the signal and the lack of
experienced personnel are among the constraints most encountered during identification of ECG signals;
also, ECG signal contains various unwanted noises that prevent the correct extraction of useful and necessary

Journal homepage: http://ijai.iaescore.com

Int J Artif Intell ISSN: 2252-8938  961

information for classification [2], [4]. Hence, finding solutions to solve these problems has become a
necessity. The field of medical engineering try the challenge by developing models capable of processing
ECG signals and dentifying any abnormalities present in the signal [5], thus, signal processing is required;
signal processing techniques analyze efficiently various kinds of signals, especially ECG signals. The aim of ECG
signal processing is to extract features to distinguish between normal signals and those representing abnormalities.
Several techniques based on signal processing have been developed over years for processing ECG
signals. Ahlstrom and Tompkins have used digital filters for real time ECG signal processing [6] to denoise
the signal and detect the QRS complex. Hargittai [7] used Savitzky-Golay Least-Squares Polynomial filters
to preserve the details of the signal, Francisco et al. [8] process ECG signal using principal component
analysis (PCA), Haque et al. [9] use adaptive filtering algorithms, while Ahmed et al. [10] use a method
based on cross correlation theory. Gustavo et al. [11] propose a comparison between several methods to
remove baseline wander. Also the choice of features and the method used for its extraction affect directly the
quality of signal characterization, some techniques are based on the extraction of morphological features such
as the detection of the QRS complex proposed by Jiapu [12], the calculation of R-R intervals and peaks
detection mentioned in the works of Shanti Chandra et al. [13] and Priyanka [14], other methods are based on
the extraction of the statistical ones or even a mixture of DWT and statistical features, Abdullah et al. [15], or
morphological and statistical features, Sahoo et al. [16].
The discrete wavelet transform (DWT) is a very powerful tool in the field of signal processing [17]-
[20], this technique gives satisfying results in the processing of the noise which affects the signal that can
allows us to reconstruct a denoised signal, also the DWT allows to extract different features that characterize
the signal. The interest of the discrete wavelet transforms (DWT) pushes us to better exploit this technique in
the processing of ECG signal [21], [22]. Also, machine learning techniques are constantly evolving; this
evolution is reflected in the use of these techniques in several areas [23], more precisely in the classification
and identification of signals. The combination of signal processing techniques and machine learning models
gives us promising results and the performance of each model changes depending on the chosen algorithm
The purpose of this modest work is to establish a characterization model of ECG signals able to
differentiate between normal and abnormal signals, the adopted model is based on the extraction of statistical
features from the approximation coefficients obtained by the wavelet decomposition of the signal and classify
these features using an SVM classifier; this model was approved after the choice of the best wavelet and the
best scale increasing it accuracy, the recordings of the used ECG signals are taken from the MIT-BIH
arrhythmia database [24], [25]. It’s composed of 48 recordings of different patients; these patients are
classified into two categories: A healthy and a sick one. After the processing of the ECG signals and the
extraction of features a classification of patients to healthy or sick ones is needed. To do this we have chosen
a support vector machine (SVM) classifier who seems most appropriate to do this task [26]-[28].

2. PRELIMINARIES
2.1. Wavelet analysis
Wavelet analysis is one of the powerful tools in signal processing, it is considered as a technique
which aims to solve the problem of non-stationary signals. This notion was introduced in the 20th century by
Haar who constructed the simplest wavelet, and then developed in the 1980s through the research work of
Mallat [18], [19], [29]. One of the powerful things of this approach is that it allows a signal to be analyzed in
time and frequency, which makes it very useful in extracting the various information contained in that signal.
The wavelet transform decomposes a signal x(t) using a series of wavelets ψa,b which derive from a
mother wavelet by a displacement in time carried out by a translation and a dilation or compression by
scaling of the mother wavelet, this series of wavelets is defined as [30]:
1 t−b
ψa,b (t) = ψ( ) (1)
√a a

where a and b are respectively the scaling and the translation coefficients ψ being the mother wavelet which
must verify the following condition [17]:
+∞
∫−∞ ψ(t)dt = 0 (2)

The wavelet transform exists in two principal forms the CWT and the discrete wavelet transform (DWT),
thus we define the continuous wavelet transform (CWT) of a signal x(t) as [18]:
1 +∞ t−b
ca,b = ∫ x(t)ψ∗ ( a )dt (3)
√a −∞

Electrocardiogram signals classification using discrete wavelet transform and... (Youssef Toulni)
962  ISSN: 2252-8938

ψ∗ represents the complex conjugate of the wavelet ψ (t).

The computation of this continuous wavelet transform seems sometimes difficult and time
consuming. The discrete wavelet transform give a solution to this problem. To do this we use an algorithm
(proposed by Mallat) [18] to calculate the discrete wavelet coefficients, this decomposition consists in
passing the signal through a succession of complementary low-pass h[n] and high-pass g[n] filters (in stages),
the low-frequency components resulting from the low-pass filters h[n] represent approximation coefficients
and those from the high frequency g [n] filter are called the detail coefficients [31] as shown in Figure 1.

2.2. Features extraction

The choice of the nature, number and method of features extraction is considered one of the decisive
steps in the characterization of a signal, since each of these parameters contain various information which
influence on the efficiency of classification. In the literature many are the features and the methods used to
characterize a signal especially for the ECG signals, there are the morphological features such as the PR, RR and
QRS intervals, which are determined basically by the detection of the QRS complex [12]-[14]. There are also
statistical features such as maximum, minimum, mean, variance and standard deviation [15], [16], [32], [33].
In this related work we will focus on the following features:
The mean value (4):
1
𝜇= ∑𝑁
𝑖=1 𝑥𝑖 (4)
𝑁

The root mean square (5):

1
𝑟𝑚𝑠 = √ ∑𝑁 2
𝑖=1 𝑥𝑖 (5)
𝑁

The variance (6):

1
𝑣= ∑𝑁
𝑖=1(𝑥𝑖 − 𝜇)
2
(6)
𝑁

Standard deviation (7):

1
𝜎 = √ ∑𝑁
𝑖=1(𝑥𝑖 − 𝜇)
2 (7)
𝑁

Kurtosis (8):

1 ∑𝑁
𝑖=1(𝑥𝑖 −𝜇)
3
𝜑= (8)
𝑁 𝜎3

Wavelet entropy (9):

1
𝐻 = − ∑𝑁 2 2
𝑖=1 𝑥𝑖 log(𝑥𝑖 ) (9)
𝑁

where N is the length of the signal x

Figure 1. Signal decomposition using DWT at 3rd level of scale

Int J Artif Intell, Vol. 10, No. 4, December 2021: 960 - 970
Int J Artif Intell ISSN: 2252-8938  963

2.3. Classification
The chosen classification technique is support vector machine (SVM), SVM is a machine learning
technique that allows binary classifications. The purpose of this method is to separate the features of the
signal into different categories [27], [28]. The purpose of this method is to separate the features of the signal
into different categories, it's based on the construction of a separation area between the different classes of
the learning set{𝑥𝑖 , 𝑦𝑖 }, with 𝑦𝑖 ∈ {−1,1} and 𝑖 = 1,2, … . , 𝑛 the lines that delimit this area are called a
hyperplane which is defined by [28], [29]:

𝑦 = 𝑤𝑥 + 𝑏 (10)

𝑤 represents a normal vector on the hyperplane and 𝑏 is the bias.

Thus, the area where we have: 𝑦𝑖 = −1 implies that 𝑤𝑥𝑖 + 𝑏 ≤ −1 (11)

Similarly, for 𝑦𝑖 = +1 we have 𝑤𝑥𝑖 + 𝑏 ≥ 1 (12)

In all cases we have 𝑦𝑖 (𝑤𝑥𝑖 + 𝑏) ≥ 1 (13)

Good separation aims at increasing the width of the margin between hyperplanes,
‖𝑤‖2
this is equivalent to minimizing knowing that𝑦𝑖 (𝑤𝑥𝑖 + 𝑏) − 1 ≥ 0. (14)
2

If we can't separate linearly these classes, then the optimization problem becomes [26]:

‖𝑤‖2
min( + 𝐶 ∑𝑛𝑖=1 𝜉𝑖 )
{ 2 (15)
𝑎𝑣𝑒𝑐 𝑦𝑖 (𝑤𝑥𝑖 + 𝑏) − 𝜉𝑖 ≥ 1

3. METHODOLOGY
In this work we want to establish a simple model that allows:
 ECG signal processing.
 Extraction of the features that characterize the signal, these features will be calculated by applying
(4)-(9).
 The classification of signals processed into normal (healthy patients) and other abnormal signals (the
sick ones) using a linear SVM classifier.
Processing techniques and extraction methods are based on discrete wavelet analysis; Figure 2
shows the process of identifying ECG signals.

3.1. Database
The ECG signals are taken from the MIT BIH data base of physionet [24], [25], this database
contains 48 recordings each recording is half hour length, the recordings are sampled at 360 Hz, in this study
we analyzed duration of one minute for each signal.

3.2. ECG signal processing

The recordings of ECG are not protected from noises; these noises cause significant perturbations
during classification and diagnosis. For ECG signals, there are different sources of noises, noises of technical
origins such powerline interference (50 or 60 Hz noise from mains supply) or bad wiring and others of
physical origins generated by the physical activity of the body such electromyographic (EMG) noise,
baseline wander (low frequencies) which are considered among the most important [34]. The steps of signal
denoising are shown in Figure 3.
In this stage of processing, we tried to eliminate these noises using the discrete wavelet
decomposition. After the decomposition of the signal into eight levels we notice that the noise (high
frequencies) is located in the detail coefficients d1 and d2. For the baseline wander we notice that the
approximation coefficient A8 corresponds to the frequency interval of these fluctuations 0-0.5 Hz [20];
Figure 4 shows detail and approximation coefficients resulting from the wavelet decomposition of recording
121 m, and Figure 5 shows part of the same recording before, as shown in Figure 5(a) and after, as shown in
Figure 5(b) denoising, the mother wavelet used in the decomposition is Symlet [35].

Electrocardiogram signals classification using discrete wavelet transform and... (Youssef Toulni)
964  ISSN: 2252-8938

Figure 2. Signal denoising process Figure 3. Diagram of the proposed method

Figure 4. Wavelet decomposition of ECG signal (record No 121) d1, d2, d3, d4, d5, d6, d7 and a8

3.3. Features extraction

After denoising the ECG signal, we decompose the denoised signal using DWT by using different
types of wavelets to analyze the signal, and then the approximation coefficients are calculated up to scale 8 as
shown in Figure 6. After the extraction of the approximation coefficients, the features mentioned above (4)-
(8) are calculated from the approximation coefficients a 1, a2, a3, a4, a5, a6, a7, a8, thus we create for each level
a data set containing the features for all the records.

3.4. Classification
The resulting dataset is introduced in a SVM classifier and it’s divided into two subsets; the first will
be served for training the model, while the second will be designed to test the performance of the model. The
choice of the training and test subsets is crucial and it can causes many problems which influence the
effectiveness of the model such as overfitting and underfitting [36]; to limit this kinds of problems we use k
flod cross validation to select this subsets [37], [38], this method consists in dividing the data set into parts of
number of k, one of these parts is chosen as a subset of test and the other 𝑘 − 1 parts as a training subset, this
process is repeated k times and each time a different part is taken to do the test. The precision of the model is
obtained by averaging the precision of each iteration [39]-[41]. In this work we took 𝑘 = 8, that is to say that
7/8 of the data set is taken for training and 1/8 is considered as a test subset. The next step will be devoted to
measuring the performance of the model for the different types of mother wavelets at different scales (up to

Int J Artif Intell, Vol. 10, No. 4, December 2021: 960 - 970
Int J Artif Intell ISSN: 2252-8938  965

the 8th scale); the accuracy (Acc), the sensitivity (Sen) and the specificity (Spe) are the quantities that inform
us about the performance of each model with [29]:
𝑇𝑁+𝑇𝑃
𝐴𝑐𝑐 = × 100 (16)
𝑇𝑃+𝑇𝑁+𝐹𝑃+𝐹𝑁

𝑇𝑃
𝑆𝑒𝑛 = × 100 (17)
𝑇𝑃+𝐹𝑁

𝑇𝑁
𝑆𝑝𝑒 = × 100 (18)
𝑇𝑁+𝐹𝑃

TP : True positive (represents normal signal who were correctly classified)

TN : True negative (represents abnormal signal who were correctly classified)
FP : False positive (represents normal signal who were incorrectly classified)
FN : False negative (represents abnormal signal who were incorrectly classified)

(a)

(b)

Figure 5. Record no 121 before and after denoising; (a) original signal and baseline wander, (b) signal after
removing noise and baseline wander

Electrocardiogram signals classification using discrete wavelet transform and... (Youssef Toulni)
966  ISSN: 2252-8938

Figure 6. Approximation coefficients a1, a2, a3, a4, a5, a6, a7, a8

4. RESULTS
Table 1 as shown in appendix, summarizes the results obtained for different wavelet families at the
first eight scales of the approximation coefficients. As we can see, the best accuracy is generally obtained at
the fourth scale. The study carried out by Siti [42] which is based on the calculation of statistical features
from the detail coefficients at the 4th, 5th and 6th and the KNN as a classification method reaches an
accuracy of 71% using the sym7 wavelet; while in this work we had reached an accuracy of 81.67% using
only the statistical features extracted from the approximation coefficient a4 for the same wavelet and an SVM
classifier; ;also in the study of Siti [42] we reached an accuracy of 85% as the best accuracy by calculating
the MFCC coefficients, while in this work we found 87.50% as better accuracy by adopting only the coif5
wavelet as the mother wavelet as shown in Table 2.

Table 2. Comparison results

Method Classifier Accuracy
Haar 68%
DWT D4, D5, D6
Siti [42] Sym 7 KNN 71%
MFCC 85%
Proposed method Sym7 81,67%
DWT A4 SVM
Coif 5 87,50%

5. CONCLUSION
In this article we focused on the classification of ECG signals, we were then able to establish a
model that allows to process and classify these signals. ECG signals were processed using DWT discrete
wavelet analysis, the discrete wavelet transform allows the calculation of approximation coefficients which
are used to extract the features for different scales, the dataset which consists of these parameters is fed into
an SVM classifier with cross validation in order to distinguish between a normal and an abnormal signal.
This model was tested with different mother wavelets and at different scales to reach an accuracy of 87.50%,
which was the best. To conclude, the choice of the mother wavelet, the scale of the decomposition as well as
the size of the training and test sets have a considerable influence on the accuracy of the model.

Int J Artif Intell, Vol. 10, No. 4, December 2021: 960 - 970
Int J Artif Intell ISSN: 2252-8938  967

APPENDIX

Table 1. Accuracy, sensitivity and specificity of the different wavelets at the first eight scales (continue)
Classifier Approximation coefficients scale level
Wavelet
performance 1 2 3 4 5 6 7 8
Accuracy 72.08 71.67 70.83 79.58 70.42 68.33 60.42 56.84
db2 Sensitivity 75.83 76.67 77.50 90.00 81.67 75.42 68.33 64.50
Specificity 68.33 66.67 64.17 69.17 59.17 60.00 52.50 50.00
Accuracy 71.25 72.92 72.92 77.92 70.42 65.42 58.75 53.51
db3 Sensitivity 74.17 76.67 79.17 86.67 79.17 74.58 69.17 53.83
Specificity 68.33 69.17 66.67 69.17 61.67 53.33 48.33 53.33
Accuracy 72.08 71.25 74.17 75.83 68.33 64.17 58.75 51.97
db4 Sensitivity 75.00 75.83 80.00 86.67 78.33 72.50 64.17 52.54
Specificity 69.17 66.67 68.33 65.00 58.33 54.17 53.33 51.67
Accuracy 72.92 70.83 71.25 77.08 68.33 65.00 60.00 39.99
db5 Sensitivity 79.17 76.67 79.17 89.17 80.00 74.17 69.17 55.53
Specificity 66.67 65.00 63.33 65.00 56.67 55.00 50.83 29.17
Accuracy 73.33 72.50 72.92 76.25 65.00 65.83 59.58 41.08
db6 Sensitivity 80.00 78.33 80.00 87.50 74.17 75.42 73.33 68.19
Specificity 66.67 66.67 65.83 65.00 55.83 53.33 45.83 18.33
Accuracy 70.83 73.33 77.50 72.08 67.08 63.33 60.00 41.68
db7 Sensitivity 78.33 79.17 85.00 81.67 75.83 75.00 74.17 69.26
Specificity 63.33 67.50 70.00 62.50 58.33 49.17 45.83 18.33
Accuracy 73.33 74.58 73.75 75.83 67.50 63.75 59.58 49.58
db8 Sensitivity 79.17 81.67 80.83 86.67 73.33 75.42 67.50 68.59
Specificity 67.50 67.50 66.67 65.00 61.67 48.33 51.67 30.00
Accuracy 72.92 71.67 72.92 77.50 70.00 67.92 60.83 54.49
sym2 Sensitivity 76.67 75.00 77.50 85.00 80.00 76.25 71.67 65.10
Specificity 69.17 68.33 68.33 70.00 60.00 57.50 50.00 45.83
Accuracy 71.67 71.25 72.08 77.08 70.83 64.17 60.83 56.30
sym3 Sensitivity 76.67 75.00 77.50 89.17 80.00 7292 72.50 58.28
Specificity 66.67 67.50 66.67 65.00 61.67 52.50 49.17 55.00
Accuracy 72.92 71.25 71.67 81.67 67.50 72.92 65.83 45.58
sym4 Sensitivity 75.83 74.17 77.50 95.00 80.83 79.58 70.83 64.00
Specificity 70.00 68.33 65.83 68.33 54.17 64.17 60.83 30.83
Accuracy 70.00 71.67 72.50 82.08 71.25 72.08 63.33 44.36
sym5 Sensitivity 75.83 75.00 75.83 93.33 85.83 81.25 68.33 59.91
Specificity 64.17 68.33 69.17 70.83 56.67 60.83 58.33 31.67
Accuracy 71.67 74.17 71.25 79.17 71.67 72.50 61.67 46.51
sym6 Sensitivity 76.67 77.50 76.67 91.67 85.00 81.25 69.17 61.18
Specificity 66.67 70.83 65.83 66.67 58.33 61.67 54.17 35.00
Accuracy 71.67 72.50 72.50 81.67 72.08 70.83 58.75 42.00
sym7 Sensitivity 76.67 76.67 75.83 91.67 85.83 81.25 66.67 58.77
Specificity 66.67 68.33 69.17 71.67 58.33 57.50 50.83 27.50
Accuracy 73.33 72.92 73.75 79.17 68.75 67.50 60.00 42.98
sym8 Sensitivity 78.33 75.83 77.50 90.00 80.00 77.50 69.17 63.21
Specificity 68.33 70.00 70.00 68.33 57.50 55.83 50.83 25.83
Accuracy 72.08 72.50 71.67 79.17 71.67 72.92 66.67 51.72
coif1 Sensitivity 75.00 75.83 75.00 88.33 82.50 80.00 70.83 65.32
Specificity 69.17 69.17 68.33 70.00 60.83 64.17 62.50 40.00
Accuracy 72.08 72.50 71.67 79.17 71.67 72.92 66.67 51.72
coif2 Sensitivity 75.00 75.83 75.00 88.33 82.50 80.00 70.83 65.32
Specificity 69.17 69.17 68.33 70.00 60.83 64.17 62.50 40.00
Accuracy 72.08 72.50 71.67 79.17 71.67 72.92 66.67 51.72
coif3 Sensitivity 75.00 75.83 75.00 88.33 82.50 80.00 70.83 65.32
Specificity 69.17 69.17 68.33 70.00 60.83 64.17 62.50 40.00
Accuracy 74.48 73.96 75.00 85.42 71.88 72.40 60.42 49.40
coif4 Sensitivity 81.25 81.25 83.33 97.92 83.33 79.69 72.92 63.96
Specificity 67.71 66.67 66.67 72.92 60.42 62.50 47.92 38.54
Accuracy 75.00 73.96 75.00 8750 73.96 73.96 55.21 42.80
coif5 Sensitivity 82.29 80.21 82.29 100.00 84.38 82.29 65.63 55.31
Specificity 67.71 67.71 67.71 75.00 63.54 63.54 44.79 33.33
Accuracy 63.75 65.83 67.50 70.42 69.17 64.58 62.92 61.70
bior1.1 Sensitivity 68.33 69.17 72.50 79.17 85.00 72.92 75.83 73.63
Specificity 59.17 62.50 62.50 61.67 53.33 53.33 50.00 50.83
Accuracy 71.67 71.25 73.33 79.17 69.58 70.42 60.83 47.44
bior1.3 Sensitivity 75.00 75.00 76.67 89.17 83.33 76.25 70.00 62.57
Specificity 68.33 67.50 70.00 69.17 55.83 63.33 51.67 35.00
Accuracy 70.42 71.25 71.25 78.75 72.50 74.17 60.42 40.99
bior1.5 Sensitivity 74.17 75.00 75.83 90.00 85.83 84.17 66.67 61.32
Specificity 66.67 67.50 66.67 67.50 59.17 61.67 54.17 22.50

Electrocardiogram signals classification using discrete wavelet transform and... (Youssef Toulni)
968  ISSN: 2252-8938

Table 1. Accuracy, sensitivity and specificity of the different wavelets at the first eight scales
Classifier Approximation coefficients scale level
Wavelet
performance 1 1 1
Accuracy 72.08 72.08 69.58 72.50 77.08 69.58 66.67 54.28
bior2.2 Sensitivity 75.00 75.00 74.17 80.00 89.17 75.83 64.17 61.56
Specificity 69.17 69.17 65.00 65.00 65.00 61.67 69.17 47.50
Accuracy 71.67 70.83 70.42 77.50 70.42 72.08 60.83 49.27
bior2.4 Sensitivity 76.67 74.17 74.17 85.83 80.83 76.25 69.17 62.49
Specificity 66.67 67.50 66.67 69.17 60.00 66.67 52.50 38.33
Accuracy 70.00 71.25 70.42 79.17 72.50 74.17 61.25 46.13
bior2.6 Sensitivity 75.00 75.83 75.83 88.33 83.33 80.83 67.50 57.49
Specificity 65.00 66.67 65.00 70.00 61.67 65.83 55.00 37.50
Accuracy 70.42 71.67 70.42 78.33 73.75 72.92 62.08 43.62
bior2.8 Sensitivity 75.00 75.83 72.50 86.67 85.00 81.25 70.00 64.06
Specificity 65.83 67.50 68.33 70.00 62.50 62.50 54.17 26.67
Accuracy 63.33 64.17 65.83 69.17 70.42 62.50 64.58 60.58
rbio1.1 Sensitivity 67.50 67.50 69.17 78.33 85.00 72.50 78.33 76.17
Specificity 59.17 60.83 62.50 60.00 55.83 49.17 50.83 46.67
Accuracy 70.00 72.50 72.92 77.08 72.92 70.42 66.67 60.36
rbio1.3 Sensitivity 75.00 75.00 78.33 85.00 82.50 77.50 80.83 75.53
Specificity 65.00 70.00 67.50 69.17 63.33 61.67 52.50 48.33
Accuracy 69.17 71.67 72.92 78.33 74.17 72.08 64.17 52.65
rbio1.5 Sensitivity 73.33 75.83 78.33 87.50 81.67 80.42 75.83 70.90
Specificity 65.00 67.50 67.50 69.17 66.67 61.67 52.50 37.50
Accuracy 72.92 72.50 75.83 76.67 70.42 70.00 57.08 52.33
rbio2.2 Sensitivity 75.83 77.50 84.17 85.83 81.67 80.83 62.50 54.11
Specificity 70.00 67.50 67.50 67.50 59.17 56.67 51.67 50.83
Accuracy 71.67 72.08 75.83 78.33 68.33 68.75 56.67 44.21
rbio2.4 Sensitivity 76.67 77.50 83.33 87.50 78.33 73.33 66.67 56.94
Specificity 66.67 66.67 68.33 69.17 58.33 62.50 46.67 34.17
Accuracy 70.42 71.25 77.50 75.42 70.00 68.75 54.17 40.63
rbio2.6 Sensitivity 75.83 76.67 85.83 83.33 79.17 75.83 65.83 56.03
Specificity 65.00 65.83 69.17 67.50 60.83 60.83 42.50 28.33
Accuracy 72.92 74.17 78.33 79.17 72.92 65.83 50.83 45.07
rbio2.8 Sensitivity 77.50 81.67 85.83 90.83 81.67 73.33 65.83 61.33
Specificity 68.33 66.67 70.83 67.50 64.17 55.83 35.83 30.83

REFERENCES
[1] “WHO,” 2017. [Online]. Available: https://www.who.int/en/newsroom/fact-sheets/detail/cardiovascular-diseases-
(cvds).
[2] R. Kher, “Signal processing techniques for removing noise from ECG signals,” Journal of Biomedical Engineering,
vol. 1 pp. 1-9, 2019, doi: 10.17303/jber.2019.3.101.
[3] S. M. Anwar, M. Gul, M. Majid, and M. Alnowami, "Arrhythmia classification of ECG signals using hybrid
features," Comp. and Mathematical Methods in Medicine, vol. 2018, pp. 1-8, 2018, doi: 10.1155/2018/1380348
[4] X.-K. Wan, H. Wu, F. Qiao, F.-C. Li, Y. Li, Y.-W. Yan, and J.-X. Wei, "Electrocardiogram baseline wander
suppression based on the combination of morphological and wavelet transformation based filtering,"
Computational and Mathematical Methods in Medicine, vol. 2019, pp. 1-7, 2019, Art. no. 7196156,
doi: 10.1155/2019/7196156
[5] A. Rahman, N. Athilah, and J. Asral, “Biomedical health monitoring system design and analysis,” Indonesian
Journal of Electrical Engineering and Computer Science (IJEECS), vol. 13, no. 3, pp. 1056-1064, 2019,
doi: 10.11591/ijeecs.v13.i3.pp1056-1064.
[6] M. L. Ahlstrom and W. J. Tompkins, "Digital filters for real-time ECG signal processing using microprocessors,"
IEEE Transactions on Biomedical Engineering, vol. 32, no. 9, pp. 708-713, 1985,
doi: 10.1109/TBME.1985.325589.
[7] S. Hargittai, “Savitzky-Golay least-squares polynomial filters in ECG signal processing,” Computers in
Cardiology, vol. 32, pp. 763-766, 2005, doi: 10.1109/CIC.2005.1588216.
[8] F. Castells, P. Laguna, L. Sörnmo, B. Andreas, and J. M. Roig, “Principal component analysis in ecg signal
processing,” EURASIP Journal on Advances in Signal Processing, vol. 2007, no. 1, 2007, Art. No. 074580,
doi: 10.1155/2007/74580
[9] Z.-ul-Haque, R. Qureshi, M. Nawaz, F. Y. Khuhawar, N. Tunio, and M. Uzair, “Analysis of ECG signal processing
and filtering algorithms,” International Journal of Advanced Computer Science and Applications (IJACSA),
vol. 10, no. 3, 2019, doi: 10.14569/IJACSA.2019.0100370
[10] S. Ahmed, A. H.-Alnaqbi, M. Al Hemairy, and M. Al Ahmad, “ECG abnormality detection algorithm,”
International Journal of Advanced Computer Science and Applications (IJACSA), vol. 9, no. 8, pp. 215-219, 2018,
doi: 10.14569/IJACSA.2018.090827
[11] G. Lenis, N. Pilia, A. Loewe, W. H. Walther Schulze, and O. Dössel, "Comparison of baseline wander removal
techniques considering the preservation of ST changes in the Ischemic ECG: A simulation study," Computational
and Mathematical Methods in Medicine, vol. 2017, pp. 1-13, 2017, Art. No. 9295029, doi: 10.1155/2017/9295029

Int J Artif Intell, Vol. 10, No. 4, December 2021: 960 - 970
Int J Artif Intell ISSN: 2252-8938  969

[12] J. Pan and W. J. Tompkins, "A real-time QRS detection algorithm," IEEE Transactions on Biomedical
Engineering, vol. BME-32, no. 3, pp. 230-236, 1985, doi: 10.1109/TBME.1985.325532.
[13] S. Chandra, A. Sharma, and G. K. Singh, “Feature extraction of ECG signal,” Journal of Medical Engineering and
Technology, vol. 42, no. 4, pp. 306-316, 2018, doi: 10.1080/03091902.2018.1492039
[14] M. Priyanka, “Detection and processing of the R Peak,” International Journal of Innovative Research in Electrical,
Electronics, Instrumentation and Control Engineering (IJIREEICE), vol. 6, no. 11, pp. 36-44, 2018,
doi: 10.17148/IJIREEICE.2018.6116.
[15] A. Chashmi and M. Amirani, “An efficient and automatic ECG arrhythmia diagnosis system using DWT and HOS
features and entropy- based feature selection procedure,” Journal of Electrical Bioimpedance, vol. 10, no. 1,
pp. 47-54, 2017, doi: 10.2478/joeb-2019-0007
[16] S. Sahoo, M. Mohanty, S. Behera, and S. K. Sabut, “ECG beat classification using empirical mode decomposition
and mixture of features,” Journal of Medical Engineering & Technology, vol. 41, no. 8, pp. 652-661, 2017,
doi: 10.1080/03091902.2017.1394386
[17] H. Feng, J. Wang, Y. Li, and J. Chen, “Wavelet theory and application summarizing,” International Conference on
Information Computing and Applications ICICA 2011, vol 7030, 2011, pp; 337-343, doi: 10.1007/978-3-642-
25255-6_43
[18] S. G. Mallat, “A theory for multiresolution signal decomposition: the wavelet representation,” IEEE Transactions
on Pattern Analysis and Machine Intelligence, vol. 11, no. 7, pp. 674-693, 1989, doi: 10.1109/34.192463.
[19] S. Mallat and W. L. Hwang, "Singularity detection and processing with wavelets," IEEE Transactions on
Information Theory, vol. 38, no. 2, pp. 617-643, 1992, doi: 10.1109/18.119727.
[20] O. Hicham, S. Eddarouich, A. Bourouhou, and M. Timouyas, “Comparison between SVM and KNN classifiers for
iris recognition using a new unsupervised neural approach in segmentation,” IAES International Journal of
Artificial Intelligence (IJ-AI), vol. 9, no 3, pp. 429-438, 2020, doi: 10.11591/ijai.v9.i3.pp429-438.
[21] I. Houamed, L. Saidi, and F. Srairi, “ECG signal denoising by fractional wavelet transform thresholding,” Research
on Biomedical Engineering, vol. 36, pp. 349-360, 2020, doi: 10.1007/s42600-020-00075-7
[22] O. El B’charri, R. Latif, K. Elmansouri, A. Abenaou, and W. Jenkalm “ECG signal performance de-noising
assessment based on threshold tuning of dual-tree wavelet transform,” Bio-Medical Engineering OnLine, vol. 16,
no. 1, 2017, Art. no. 26, doi: 10.1186/s12938-017-0315-1
[23] A. Aliyu, M. Musa, and S. Usman, “Machine learning for plant disease detection: An investigative comparison
between support vector machine and deep learning,” IAES International Journal of Artificial Intelligence (IJ-AI),
vol. 9, no. 3, pp. 670-683, 2020, doi: 10.11591/ijai.v9.i4.pp670-683.
[24] B. G. Moody and G. R. Mark, ‘The impact of the MIT-BIH arrhythmia database,” IEEE Engineering in Medicine
and Biology Magazine, vol. 20, no. 3, pp. 45-50, 2001, doi: 10.1109/51.932724.
[25] A. Goldberger, L. Amaral, L. Glass, J. Hausdorff, C. P. Ivanov, R. Mark, and E. H. Stanley, “PhysioBank,
PhysioToolkit, and PhysioNet: Components of a new research resource for complex physiologic signals,”
Circulation, vol. 101, no. 23, pp. e215-e220, 2000, doi: 10.1161/01.cir.101.23.e215.
[26] S. Ibrahim, N. A. Zulkifli, N. Sabri, and A. Amilah, M. R. Mohd Noordin, “Rice grain classification using multi-
class support vector machine (SVM),” IAES International Journal of Artificial Intelligence (IJ-AI), vol. 8, no. 3,
pp. 215-220, 2019, doi: 10.11591/ijai.v8.i3.pp215-220
[27] A. Shmilovici, O. Maimon, and L. Rokach, “Data mining and knowledge discovery handbook,” Springer, Boston,
MA, 2009, doi: 10.1007/978-0-387-09823-4_12.
[28] K. S. Parikh and T. P. Shah, "Support vector machine-a large margin classifier to diagnose skin illnesses," Procedia
Technology, vol. 23, pp. 369-375, 2016, doi: 10.1016/j.protcy.2016.03.039.
[29] T. B. Drissi, Z. Soumaya, N. Benayad, and A. Abdelkrim, “Diagnosis of Parkinson’s disease based on wavelet
transform and mel frequency cepstral coefficients,” International Journal of Advanced Computer Science and
Applications, vol. 10, no. 3, 2019, doi: 10.14569/IJACSA.2019.0100315.
[30] I. Hammami, L. Salhi, and S. Labidi, “Voice pathologies classification and detection using EMD-DWT analysis
based on higher order statistic features,” IRBM, vol. 41, no. 3, pp. 161-171, 2020,
doi: 10.1016/j.irbm.2019.11.004.
[31] K. Surekha and B. P. Patil, “ECG Signal compression using the high frequency components of wavelet transform,”
International Journal of Advanced Computer Science and Applications, vol. 7, no. 3, pp. 311-315, 2016,
doi: 10.14569/IJACSA.2016.070344.
[32] A. Darmawahyuni et al., “Delineation of electrocardiogram morphologies by using discrete wavelet transforms,”
Indonesian Journal of Electrical Engineering and Computer Science (IJEECS), vol. 22, no. 1, pp. 159-167,
doi: 10.11591/ijeecs.v22.i1.pp159-167.
[33] E. A. Nidhal and F. Zahraa, “Detection and recognition of brain tumor based on DWT, PCA and ANN,” Indonesian
Journal of Electrical Engineering and Computer Science (IJEECS), vol. 18, no. 1, pp. 56-63, 2020,
doi: 10.11591/ijeecs.v18.i1.pp56-63.
[34] A. Subasi, “Chapter 4-feature extraction and dimension reduction, practical guide for biomedical signals analysis
using machine learning techniques,” Academic Press, 2019, pp. 193-275, doi: 10.1016/B978-0-12-817444-
9.00004-0.
[35] S. Seung Won, K. S. Kim, C. G. Song, J. W. Lee, J. H. Kim, and G. W. Jeung, “Removal of baseline wandering in
ECG signal by improved detrending method,” Bio-medical Materials and Engineering, vol. 26, no. 1, pp. S1087-
S1093, 2015, doi: 10.3233/BME-151405.
[36] B. Krishna, “Research of machine learning algorithms using K-Fold cross validation,” International Journal of
Engineering and Advanced Technology, vol. 8, no. 6S, pp. 215-218, 2021.

Electrocardiogram signals classification using discrete wavelet transform and... (Youssef Toulni)
970  ISSN: 2252-8938

[37] G. Garg, V. Singh, J. R. P. Gupta, and A. P. Mittal, "Optimal algorithm for ECG denoising using discrete wavelet
transforms," 2010 IEEE International Conference on Computational Intelligence and Computing Research,
Coimbatore, 2010, pp. 1-4, doi: 10.1109/ICCIC.2010.5705839.
[38] W. N. W. Md Adnan, N. Y. Dahlan, and I. Musirin, “Development of option c measurement and verification model
using hybrid artificial neural network-cross validationtechnique to quantify saving,” IAES International Journal of
Artificial Intelligence (IJ-AI), vol. 9, no. 1, pp. 25-32, 2020, doi: 10.11591/ijai.v9.i1.pp25-32.
[39] N. Seman, and N. A. Razmi, “Machine learning-based technique for big data sentiments extraction,” IAES
International Journal of Artificial Intelligence (IJ-AI), vol. 9, no. 3, pp. 473-479, 2020,
doi: 10.11591/ijai.v9.i3.pp473-479.
[40] X. Zheng, C. Yuxin, L. Zhonghao, H. Ming, and H. Jianjun, “Evaluating explorative prediction power of machine
learning algorithms for materials discovery using k-fold forward cross-validation,” Computational Materials
Science, vol. 171, 2020, Art. No. 109203, doi: 10.1016/j.commatsci.2019.109203.
[41] S. Yadav and S. Shukla, "Analysis of k-Fold Cross-validation over hold-out validation on colossal datasets for
quality classification," 2016 IEEE 6th International Conference on Advanced Computing (IACC), Bhimavaram,
2016, pp. 78-83, doi: 10.1109/IACC.2016.25.
[42] S. A. Alodia Yusuf and R. Hidayat, "MFCC feature extraction and KNN classification in ECG signals," 2019 6th
International Conference on Information Technology, Computer and Electrical Engineering (ICITACEE),
Semarang, Indonesia, 2019, pp. 1-5, doi: 10.1109/ICITACEE.2019.8904285.

BIOGRAPHIES OF AUTHORS

Youssef Toulni was born in Casablanca, Morocco in 1983. Graduated with a Masters in
computer science and scientific instrumentation in high energy physics from the Ain Chok
Faculty of Sciences. Hassan II University - Casablanca, Morocco, in 2019, he is a student-
researcher at the Industrial, Computer and Logistics Engineering Research Laboratory
(GITIL). Faculty of Sciences Ain Chok, Hassan II University - Casablanca, Morocco. He is
interested in the processing of biomedical signals.

Taoufiq Belhoussine Drissi was born in Oujda, Morocco in 1978 received the Ph.D. degree in
acoustics in 2009 at the university of le Havne (France) Since 2011 he has been an assistant
professor at the sciences faculty of Ain chock university Hassan II, Casablanca. His scientific
interest lies in the research of nondestructive testing and the signal treatment.

Benayad Nsiri held MBI degree in computer sciences from Telecom Bretagne, in 2005, and
Ph.D. degree in signal processing from Telecom Bretagne, in 2004. He received D.E.A
(French equivalent of M.Sc. degree) in electronics from the Occidental Bretagne University, in
2000. Currently, he is a Full Professor at the National School of Arts and Crafts of Rabat
(ENSAM), Mohammed V University; a member in Research Center STIS, M2CS, Mohammed
V University; and a member associate in Researcher, Industrial Engineering, data processing
and logistic Laboratory, Hassan II University. He was a Professor in the Faculty of Sciences
Ain Chock, Hassan II University. Benayad NSIRI has advised and co-advised more than 12
Ph.D. theses, contributed to more than 80 articles in regional and international conferences and
journals. His research interests include but not restricted to computer science,
telecommunication, signal and image processing, adaptive techniques, blind deconvolution,
MCMC methods, seismic data and higher order statistics.

Int J Artif Intell, Vol. 10, No. 4, December 2021: 960 - 970