https://www.irjet.net/archives/V9/i2/IRJET-V9I2156.pdf

1. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 02 | Feb 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 908
A Review on ECG -Signal Classification of Scalogram Snap shots the use
of Convolutional Neural Network and Continuous Wavelet Transform
Padmapriya Veerappan1*, Murugesan Manikkampatti Palanisamy2, M Depha3
1&3Department of Electronics Communication Engineering Erode Sengunthar Engineering College, Erode,
Tamil Nadu, India
2Deparment of Food Technology, Excel Engineering College, Namakkal, Tamil Nadu, India
---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract – Electrocardiogram (ECG)is broadly applied tool
for the identification of cardiovascular problems. Due to the
increased growth of populace, computer-primarily based
computerized ECG analyzer has won extra significance.
Arrhythmia should be detected early and treated effectively,
reducing the number of deaths caused by cardiovascular
disease (CVD). The diagnosis is normally made by testing the
electrocardiogram (ECG) beat by beat in clinical practice,
but this is time-consuming and laborious. Previous research
discuses in ECG classification method primarily based on
continuous Wavelet transform (CWT) and Convolutional
Neural Networks (CNN).The study is performed which
Proposed system achieves an average output of specificity,
sensitivity, and accuracy respectively, While tested inside the
three magnificence of ECG database with the aid of the
usage of the Morlet wavelet.
Key Words: Electrocardiogram, Cardiovascular disease,
Continuous Wavelet Transform, Convolutional Neural
Networks, Morlet wavelet
1. INTRODUCTION
Electro cardiogram (ECG) sign is a graphical illustration of
cardiac pastime and it used to measure the various cardiac
sicknesses and abnormalities present in heart. ECG
indicators are composed of P wave, QRS complicated, T
wave and any deviation in these parameters imply
abnormalities present in coronary heart. as a way to
report an ECG signal, electrodes (transducers) are placed
at specific positions at the human body. the automated
prognosis of coronary heart ailment is based heavily at the
class of electrocardiogram (ECG) indicators. It is
historically divided into two steps: feature extraction and
pattern category. It’s been shown that a deep neural
network educated on a big volume of records can carry out
function extraction at once from the statistics and hit upon
cardiac arrhythmias higher than expert cardiologists, way
to recent tendencies in artificial intelligence. Bellow a
comprehensive survey of Accuracy for Convolutional
Neural Networks (CNN) detection monitored arrhythmia
graphical representation of electro aerobic picture signals
are summarized and the path of reduction in identification
time and growth the accuracy rate model.
A non-stop Wavelet Transform (CWT) and Convolutional
Neural community-based totally technique used for the
ECG class. The 2nd-scalogram composed of the above
time-frequency additives is decomposed the usage of CWT
to get right of entry to numerous time-frequency additives,
and CNN is used to separate features from the second-
scalogram composed of the above time-frequency
components. However, marking ECG heartbeats is each
luxurious and time eating [1]. A unique deep-learning
technique for ECG type based totally on adversarial
domain version, improves the phenomenon of different
information distribution due to person variations, and
enhances the classification accuracy of move-area ECG
indicators with exceptional statistics distributions [2].
2. LITERATURE SURVEY
The look at shows a -dimensional deep convolutional
neural network approach for ECG arrhythmia
classification. ECG time domain signals are first
transformed into time-frequency spectrograms the use of
a short-time Fourier rework. Following that, the second-
CNN became used to perceive and classify the ECG
arrhythmia sorts the use of the spectrograms of the
arrhythmia kinds as facts [3]
The electrocardiography (ECG) arrhythmias have been
labeled in this examine the usage of a proposed system
that depended on deep neural networks to extract records.
As inputs to deep convolutional neural networks, the
cautioned techniques use a huge variety of uncooked ECG
time-collection statistics and ECG signal spectrograms
(CNN).the primary technique makes use of ECG time-
series signals immediately as enter to CNN, whilst the
second method converts ECG indicators into time-
frequency area matrices and sends them to CNN. For
immediate and efficient operation of the CNN algorithm,
the most appropriate parameters together with the variety
of layers, size, and quantity of filters are heuristically
optimized. Using deep gaining knowledge of algorithms
rather than traditional function extraction techniques, the
proposed gadget confirmed a excessive type overall
performance for time-collection information and
spectrograms. The common sensitivity, precision, and
accuracy values are used to evaluate efficiency [4].
Electrocardiograms (ECGs) are classified a good way to
resource in the medical prognosis modern-day heart
sickness. With a convolutional neural network (faster R-

2. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 02 | Feb 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 909
CNN) algorithm, this paper proposes an efficient method
development and implementation for ECG class based
totally on quicker regions. In this test, the original one-
dimensional ECG indicators consist of preprocessed
patient ECG indicators in addition to a few ECG recordings
from the MIT-BIH database. For the experimental
schooling and take a look at sets, each ECG beat of one-
dimensional ECG indicators changed into transformed
right into a -dimensional image [5].
Primarily based on numerous ECG functions, this study
proposed an ECG (Electrocardiogram) classification
technique using system today's. An electrocardiogram
(ECG) is a signal that facts the coronary heart's electrical
interest. The proposed approach is applied on the Apache
Spark framework the usage of ML-libs and the Scala
programming language; MLlib is Apache Spark's scalable
device today's library. The most hard component latest
ECG classification is managing irregularities inside the
signals, which are crucial for detecting patient reputation
[6].
Researchers gift the ECG signal preprocessing, heartbeat
segmentation strategies, characteristic description
techniques, and state-of-the-art algorithms used in this
work to survey the modern techniques ultra-modern ECG-
based totally automated abnormalities heartbeat category.
We also pass over latest the databases which have been
used to assess techniques that have been recommended by
way of a well-known evolved by way of the association for
the development latest medical Instrumentation [7].
The main intention present day this challenge is to
develop a deep studying-primarily based method so that it
will cutting-edge they want for manual feature
identification. We used five, 655 unmarried-lead ECG
recordings to teach a pre-educated convolutional neural
network (CNN) called AlexNet. to begin with, we created a
spectrogram for all 30s indicators and used continuous
Wavelet remodel to convert them to RGB images. We first
extracted a spectrogram for all 30s indicators and used
non-stop Wavelet transform (CWT) to convert them to
RGB snap shots, which we then fed to AlexNet and
educated with a few modifications in specifications [8].
This paper proposes a singular hostile area model-based
deep-gaining knowledge of technique for ECG category
that addresses the issue of inadequately labelled training
samples, improves the phenomenon of different statistics
distribution because of person variations, and improves
the classification accuracy of go-domain ECG alerts with
specific information distributions. The time features and
the deep-mastering extraction capabilities are
concatenated on the completely linked layer to decorate
feature range [9].
We suggest a two-dimensional (2-D) convolutional neural
network (CNN) version for categorizing ECG statistics into
8 lessons: everyday beat, untimely ventricular contraction
beat, paced beat, proper bundle branch block beat, left
package branch block beat, atrial untimely contraction
beat, ventricular flutter wave beat, and ventricular escape
beat on this take a look at. Short-time Fourier rework is
used to transform one-dimensional ECG time series
records into -dimensional spectrograms. The 2-D CNN
version, which has 4 convolutional layers and four pooling
layers, is meant to extract sturdy characteristics from
enter spectrograms [10].
2.1. Motivation of the work
The detection of ECG arrhythmias is important for the
treatment of sufferers and the early prognosis of cardiac
sickness. Medical doctors discover it tough to examine
huge ECG recordings in a quick quantity of time, and the
human eye is sick-ready to discover morphological
versions in ECG alerts, necessitating the usage of a
computer-aided diagnostic device.
2.2. Objective of the work
Accuracy for Convolutional Neural Networks (CNN)
detection monitored arrhythmia graphical representation
of electro aerobic photo signals. To reduce the
identification time and boom the accuracy rate for green
model to classify the Electro cardiogram (ECG) arrythmias.
3. SOFTWARE USAGE & DATABASE
3.1. MATLAB R2020b
Math Works is the leading developer of mathematical
computing software program. MATLAB, the language of
engineers and scientists, is programming surroundings for
set of rules improvement, statistics analysis, visualization,
and numeric computation. Simulink is a block diagram
surroundings for simulation and version-based totally
layout of multi domain and embedded engineering
structures. Engineers and scientists international rely
upon these product families to accelerate the tempo of
discovery, innovation, and improvement in automotive,
aerospace, electronics, economic services, biotech-
pharmaceutical, and different industries. R2020b
introduces new products that build on AI capabilities,
speed up self sufficient structures improvement, and boost
up advent of three-D scenes for automatic driving
simulation.
3.2. PHYSIO BANK ATM
Physio Bank is a large and growing archive of well-
characterized virtual recordings of physiological alerts
and related records to be used by way of the biomedical
studies community. The ATM gives a toolbox for facts
exploration and export, however particularly gradual
waveform show in case you need only a small amount of

3. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 02 | Feb 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 910
facts for observe, the Physio bank ATM allows you to gain
as much as a 100,000 samples of digitized indicators of
your desire in textual content layout, and it similarly
converts any amount of annotations into text format.[10-
12]. The benchmark dataset for comparing the counseled
method inside the challenge is the MIT-BIH arrhythmia
database. The ECG sign Set is made from 162 ECG
recordings taken from the subsequent PhysioNet
databases [11-12]. For CNN schooling procedure the
database pre-processed. Every recording includes 65,536
samples; it may be broken into small alerts of period to
increase the scale of database to make it suitable to teach a
CNN. MIT-BIH Arrhythmia Database– ninety six
Recordings, MIT-BIH NSR Database – 30 Recordings,
BIDMC CHF Database – 36 Recordings. Observe can
suggest work includes 5 steps and the steps are prepared
as follows:
Step 1: provides literature survey of various classification
techniques.
Step 2: Explains the Proposed methodology
Step 3: offers details of software utilization & Database
Step 4: speak about the results acquired
Step five: Concludes the record
4. CONCLUSIONS
From the overview have been concludes that the Least
mean square (LMS) and Recursive least Squares (RLS)
algorithms get rid of numerous noises from ECG alerts
efficiently. Generally RLS filtered output has the higher
overall performance than the LMS set of algorithm. The
proposed technique has the decreased imply mean square
error value (MSEV) and SNR development than the other
two adaptive algorithms. Through using the pretrained
neural network the ECG sign class becomes achieved. CNN
had a major advantage in that it brought the idea of the
usage of second images in preference to manually
extracting and choosing functions. Furthermore,
improvements in the area of deep image category aided
the task. The time-frequency scalogram of every ECG
recording is extracted the usage of CWT. To teach the
picture in AlexNet, the scalograms are translated into RGB
pictures with the perfect dimensions. Records are
subjected into CNN (AlexNet), that's then qualified. Then,
for our intended reason, a CNN version is advanced, which
has yielded a positive end result whilst trying out the test
dataset. While comparing the two special wavelet, Morlet
wavelet give the best category accuracy.
Nomenclature
CNN – Convolutional Neural Networks
CVD – Cardiovascular Disease
CVD – Cardiovascular Disease
CWT – Continuous Wavelet Transform
ECG – Electrocardiogram
LMS – Least Mean Square Algorithms
RLS – Recursive Least Squares Algorithms
RMSEV – Reduced Mean Square Error Value
ACKNOWLEDGEMENT
This have a look at became completed with usage of the
laboratory facilities in Erode Sengunthar Engineering
College and Excel Engineering College. The corresponding
author would really like well known and thank to her dad
and mom and brother Dr. P. Selvarasu, PG assist Zoology,
higher Secondary school, Vellore for his support.
REFERENCES
[1] T. Wang, C. Lu, Y. Sun, M. Yang, C. Liu, C. Ou.
“Automatic ECG Classification Using Continuous
Wavelet Transform and Convolutional Neural
Network” Entropy (Basel). 2021 Jan 18;23(1):119.
[2] L. Niu, C. Chen, H. Liu, S. Zhou, M. Shu. “A Deep-
Learning Approach to ECG Classification Based on
Adversarial Domain Adaptation” Healthcare (Basel).
2020 Oct 27;8(4):437.
[3] J. Huang, B. Chen, B. Yao and W. He, "ECG Arrhythmia
Classification Using STFT-Based Spectrogram and
Convolutional Neural Network," in IEEE Access,
vol. 7, pp. 92871-92880, 2019,
[4] S. Y. Şen and N. Özkurt, "ECG Arrhythmia Classification
By Using Convolutional Neural Network And
Spectrogram," 2019 Innovations in Intelligent Systems
and Applications Conference (ASYU), 2019, pp. 1-6,
doi: 10.1109/ASYU48272.2019.8946417.
[5] Ji Y, Zhang S, Xiao W. Electrocardiogram Classification
Based on Faster Regions with Convolutional Neural
Network. Sensors (Basel). 2019 Jun 5;19(11):2558.
doi: 10.3390/s19112558. PMID: 31195603; PMCID:
PMC6603727.
[6] F.I. Alarsan, M. Younes, “Analysis and classification of
heart diseases using heartbeat features and machine
learning algorithms” Journal of Big Data 6, 81 (2019).
https://doi.org/10.1186/s40537-019-0244-x
[7] E.J. Luz, W.R. Schwartz, G. Cámara-Chávez, D. Menotti.
“ECG-based heartbeat classification for arrhythmia

4. International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 09 Issue: 02 | Feb 2022 www.irjet.net p-ISSN: 2395-0072
© 2022, IRJET | Impact Factor value: 7.529 | ISO 9001:2008 Certified Journal | Page 911
detection: A survey. Comput Methods Programs”
Biomed. 2016 Apr; 127:144-64.
doi: 10.1016/j.cmpb.2015.12.008. Epub 2015 Dec 30.
PMID: 26775139.
[8] F. R. Mashrur, A. Dutta Roy and D. K. Saha, "Automatic
Identification of Arrhythmia from ECG Using AlexNet
Convolutional Neural Network," 2019 4th
International Conference on Electrical Information
and Communication Technology (EICT), 2019, pp. 1-5,
doi: 10.1109/EICT48899.2019.9068806.
[9] A. Ullah, S. Anwar, M. Bilal, R. Mehmood, M. Raja.
“Classification of Arrhythmia by Using Deep Learning
with 2-D ECG Spectral Image Representation” Remote
Sensing.12. 1685. 10.3390/rs12101685.
[10] M.F. Guo, X.D. Zeng, D. Y. Chen, N.C. Yang. Deep-
learning-based earth fault detection using continuous
wavelet transform and convolutional neural network
in resonant grounding distribution systems. IEEE
Sens. J. 2017, 18, 1291–1300.
[11] S. C. Olhede, A. T. Walden, "Generalized Morse
wavelets," in IEEE Transactions on Signal Processing,
vol. 50, no. 11, pp. 2661-2670, Nov. 2002, doi:
10.1109/TSP.2002.804066.
[12] Addison, P.S. (2016). The Illustrated Wavelet
Transform Handbook: Introductory Theory and
Applications in Science, Engineering, Medicine and
Finance, Second Edition (2nd ed.). CRC Press.
BIOGRAPHIES
Mrs. V. Padmapriya is a PG Scholar
(Applied Electronics) of the Department
of ECE at Erode Sengunthar Engineering
College, Tamil Nadu, and India. Her BE
ECE at Thanthai Periyar Government
Institute of Technology, Vellore under
Anna University Chennai in 2018. She has
published 2 Book chapter from Wildest
first open access publisher Intechopen
Limited, Prinnces Gate Court, London,
UK. She has published research article in
reputed International Journal.
Dr. M P Murugesan (Chemical
Engineering) is presently working as
Associate Professor in the Dept of Food
Tech at Excel Engineering College,
Tamilnadu, India. He has good research
experience in Novel treatment
Techniques for E- waste management. In
his more than 10 years of teaching
experience, he has published more than
15 scientific papers in peer reviewed
international journals and Book chapter
at Wildest first open access publisher
Intechopen Limited, London, UK.. He got
the International Best Researcher award
from ISSN (IIRA-22) and is member of
professional organizations.
Dr. M. Dhipa working as Assistant
Professor in the Department of ECE at
Erode Sengunthar Engineering College,
TamilNadu, India. She completed B.E
(EIE) in Easwari Engineering College,
Chennai in 2004 and M.E (Applied
Electronics) in K.S.R.College of
Technology, Anna University, in 2006.
She pursued Ph.D. under Anna
University. She is having 12 years of
teaching experience. She has published
about 8 papers in various International
Journals. Her area of interest includes
Heterogeneous Networks, Mobile
Computing and IoT