The document presents an automated algorithm for detecting electromyogram (EMG) noise in electrocardiogram (ECG) data using morphological operations. The algorithm efficiently filters out EMG noise to ensure clear ECG signals, utilizing techniques such as moving averaging and thresholding, and achieves a 100% detection rate with low computational complexity. This approach is suitable for application in large medical data collections and cloud-based medical systems.

1. Berk SOYSAL 180206053
1
EE434 Biomedical Signal Processing Project
Article: An Algorithm for EMG Noise Detection in Large ECG Data
(P Raphisak, SC Schuckers, A de Jongh Curry)
Abstract
Electrocardiogram recordings (ECG)
are valuable for researchers. However;
some sections of the recorded ECG may
be corrupted by electromyogram (EMG)
noise from muscle.
An electromyogram (EMG) is a test
that is used to record the electrical activity
of muscles. When muscles are active,
they produce an electrical current. This
current is usually proportional to the level
of the muscle activity.In order to obtain
clear ECG signal, EMG noise needs to be
detected and filtered before performing
data processing.
In this project, an automated
algorithm for detecting EMG noise in
large ECG data is presented. The
algorithm extracts EMG artifact from the
ECG by using morphological operations.
1. Introduction
Electrocardiography (ECG ) is the
recording of the electrical activity of the
heart. An ECG is used to measure the
heart’s electrical conduction system.
In this project, a fast aIgorithm for
detecting EMG noise is implemented
which uses morphological filters. It has
low computational complexity therefore it
can be applied in massive ECG
collections.
2. Methods
2.1 Morphological Operations
Morphological filters have been
widely used in the field of image
processing. It has applications in filtering
unwanted shapes of the signal while
leaving the other parts of signal
unchanged. The fundamental operations
of morphological filters is accomplished
by erosion and dilation operations.
The erosion of the binary image A by the
structuring element B is defined by:
The erosion of the dark-blue square by a
disk, resulting in the light-blue square.
The dilation of A by the structuring
element B is defined by:
The dilation of the dark-blue square by a
disk, resulting in the light-blue square with
rounded corners.
Opening and closing are two operations,
defined in term of erosion and dilation.
Opening is defined as erosion followed by
dilation while closing is dilation followed
by erosion.
The morphological filter is constructed
based on these operations and illustrated
in Fig. 1.
Fig. 1. Diagram of morphological filter

2. Berk SOYSAL 180206053
2
2.2 Algorithm
EMG is produced by muscle
electrical activity. In recorded ECG; EMG
interference appears as rapid fluctuations
which vary faster than ECG waves.
EMG noise in the ECG can be
detected by the developed EMG
detection algorithm. The algorithm is
described & follows.
Fig. 2. EMG Detection Algorithm
1. Extracting EMG noise: Impulsive
noise, such as EMG can be separated
from ECG by using morphological filter
with a dome-like structuring element
which is smaller than ECG waves. The
EMG noise can be retrieved back by
subtracting the output of the
morphological filter from the input ECG.
2. Suppressing QRS: The output from
the EMG extraction morphological filter
contains EMG noise and partial QRS
complexes which need to be suppressed.
By using the beat detection algorithm, the
QRS complexes can be located and
suppressed by squaring.
3. Moving Averaging: A moving average
filter is used to obtain the envelope of the
EMG noise.
4. Thresholding: In order to remove any
components of ECG signal thresholding
is used.
3. Results
Fig. 3. Extracting EMG noise
Fig. 4. Suppressing QRS & M.Averaging
Fig. 5. Thresholding & EMG Detection

3. Berk SOYSAL 180206053
3
%% Berk Soysal EE434
close all;
% EMG signal construction
r = 0.03*randn(length(newdata),1);
t = 0.019*randn(length(newdata),1);
r(500:2000)=0;
t(1:2000)=0;
r(3000:4170)=0;
t(3500:4170)=0;
x=newdata+r+t; % ECG Signal Corrupted by
EMG Signal
subplot(4,1,1);plot(x); title(' ECG Signal
Corrupted by EMG Signal' );
SE=[0, 1, 0;1,1,1;0,1,0 ]; % Structuring Element
for Morphological Operations
x2 = imopen(x,SE); % Opening
x3 = imclose(x2,SE); % Closing
subplot(4,1,2);plot(x3); title(' ECG Signal After
Morphological Operations' );
% QRS Detector
x4 = diff(x3);
x4= diff(x4);
x4=x4.*x4;
subplot(4,1,3);plot(x4); title('QRS Compressed' );
for i=1:length(x4)
if ( x4(i) > 0.001 ) x5(i) =1;
else x5(i)=0;
end
end
%Moving Average Filter
movav=ones(200,1)/200;
x6=conv(movav,x5);
for i=1:length(x6)
if ( x6(i) > 0.001 ) x7(i) =0.3;
else x7(i)=0;
end
end
subplot(4,1,4); plot(x7); axis([0 length(x7) 0
0.5]);title('Detected EMG Noise Regions');
figure;plot(x); hold on
plot(x7,'r'); axis([0 length(x7) -0.2 0.5]);
title('Detected EMG Noise Regions');
4.Conclusion
In this project, a fast EMG detection
algorithm was explained in detail. EMG is
extracted by two steps. First, a
morphological filtering was operated,
second EMG and QRS complexes were
extracted. Then, QRS complexes were
detected and suppressed. EMG was
detected by setting a threshold on its
moving average. Moving average size
and threshold value are 200 seconds and
0.001, respectively.
The algorithm achieved 100%
detection rate on the sample ECG signal.
Computational complexity of the
algorithm is very low therefore it would be
very convenient in medical cloud
applications.
5.References
[1] Sun Y, Ramires FJ, Weber KT. Fibrosis of
atria and great
vessels in response to angiotensin II or
aldosterone infusion.
Cardiovascular Research July 1997;35( 1): 138-
147.
[2] Raphisak P, de Jongh Curry A, Malkin RA,
Schuckers
SC. Heart rate variability in rats with aldosterone-
induced
chronic heart failure. IEEE Engineering Medicine
and Biology Society Annual Conference
Proceedings September
Impulsive noise suppression and
background normalimtion of electrocardiogram
signals using morphological operators. EEE
Transactions on Biomedical Engineering Febuary
1989;2(2):262-273.
[3] Chu CH, Delp EJ.
1990; 167-170.
[4] Wang D, He DC. A fast implementation of I-d
grayscale
morphological filters. IEEE Transactions on
Circuits and
Systems TI September 1994;41(9):634436.
[5] MacDonald R, Jenkins J, Anbaecher R,
Throne R. A
software trigger for intracardiac waveform
detection with
automatic threshold adjustment. Computers in
Cardiology
2003;1:228-231.

4. Berk SOYSAL 180206053
4