Journal of

Biomedical Engineering and Research

Research Open Access

Signal Processing Techniques for Removing Noise from ECG Signals

Rahul Kher*
G H Patel College of Engineering & Technology, Vallabh Vidyanagar, Gujarat, India

*Corresponding author: Dr. Rahul Kher, G H Patel College of Engineering & Technology, Vallabh Vidyanagar, Gujarat, India,
Email: [email protected]
Received Date: February 15, 2019; Accepted Date: March 10, 2019; Published Date: March 12, 2019

Citation: Rahul Kher (2019) Signal Processing Techniques for Removing Noise from ECG Signals. J Biomed Eng 1: 1-9.

Abstract
The electrocardiogram (ECG) signals contain many types of noises- baseline wander, powerline interference, electromyo-
graphic (EMG) noise, electrode motion artifact noise. Baseline wander is a low-frequency noise of around 0.5 to 0.6 Hz. To
remove it, a high-pass filter of cut-off frequency 0.5 to 0.6 Hz can be used. Powerline interference (50 or 60 Hz noise from
mains supply) can be removed by using a notch filter of 50 or 60 Hz cut-off frequency. EMG noise is a high frequency noise of
above 100 Hz and hence may be removed by a low-pass filter of an appropriate cut-off frequency. Electrode motion artifacts
can be suppressed by minimizing the movements made by the subject. The chapter introduces the types of common noise
sources in ECG signals and simple signal processing techniques for removing them, and also presents a section of Matlab
code for the techniques described.

Keywords: Baseline wander, powerline interference, electrode motion artifacts, EMG noise, low-pass filter, high-pass filter,
notch filter

Introduction
Electrocardiogram (ECG) is a signal that describes
the electrical activity of the heart. The ECG signal is
generated by contraction (depolarization) and relaxation
(repolarization) of atrial and ventricular muscles of the
heart. The ECG signal contains- a P wave (due to atrial
depolarization), a QRS complex (due to atrial repolarization
and ventricular depolarization) and a T wave (due to
ventricular repolarization). A typical ECG signal of a normal
subject is shown in (figure 1). In order to record an ECG signal,
electrodes (transducers) are placed at specific positions on the
human body. Artifacts (noise) are the unwanted signals that
are merged with ECG signal and sometimes create obstacles
for the physicians from making a true diagnosis. Hence, it is
necessary to remove them from ECG signals using proper
signal processing methods. There are mainly four types
Figure 1: An ECG signal with typical time intervals
of artifacts encountered in ECG signals: baseline wander,
powerline interference, EMG noise and electrode motion
artifacts. They are discussed briefly below.

©2019 The Authors. Published by the JScholar under the terms of the Creative
Commons Attribution License http://creativecommons.org/licenses/by/3.0/,
which [email protected]
[email protected]

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101

 2

1.1 Baseline Wander

Baseline wander or baseline drift is the effect where the base
axis (x-axis) of a signal appears to ‘wander’ or move up and
down rather than be straight. This causes the entire signal
to shift from its normal base. In ECG signal, the baseline
wander is caused due to improper electrodes (electrode-skin
impedance), patient’s movement and breathing (respiration).
Figure 3: ECG affected by powerline (50/ 60 Hz) interference
Figure 2 shows a typical ECG signal affected by baseline
[2]
wander.

The frequency content of the baseline wander is in the range

1.3 EMG Noise
of 0.5 Hz. However, increased movement of the body during
exercise or stress test increase the frequency content of The presence of muscle noise represents a major problem in
baseline wander. Since the baseline signal is a low frequency many ECG applications, especially in recordings acquired
signal therefore Finite Impulse Response (FIR) high-pass zero during exercise, since low amplitude waveforms may
phase forward-backward filtering with a cut-off frequency of become completely obscured. Muscle noise is, in contrast
0.5 Hz to estimate and remove the baseline in the ECG signal to baseline wander and 50/60 Hz interference, not removed
can be used [3]. by narrowband filtering, but presents a much more difficult
filtering problem since the spectral content of muscle activity
considerably overlaps that of the PQRST complex. Since the
ECG is a repetitive signal, techniques can be used to reduce
muscle noise in a way similar to the processing of evoked
potentials. Successful noise reduction by ensemble averaging
is, however, restricted to one particular QRS morphology at
a time and requires that several beats be available. Hence,
there is still a need to develop signal processing techniques
which can reduce the influence of muscle noise [4]. Figure
below shows an ECG signal interfered by an EMG noise.
Figure 2: An ECG Signal with baseline wander (drift) [1]

1.2 Powerline Interference

Electromagnetic fields caused by a powerline represent a

common noise source in the ECG, as well as to any other
bioelectrical signal recorded from the body surface. Such
noise is characterized by 50 or 60 Hz sinusoidal interference, Figure 4: ECG signal with electromyographic (EMG) noise
possibly accompanied by a number of harmonics. Such
narrowband noise renders the analysis and interpretation of 1.4 Electrode Motion Artifacts
the ECG more difficult, since the delineation of low-amplitude
Electrode motion artifacts are mainly caused by skin
waveforms becomes unreliable and spurious waveforms may
stretching which alters the impedance of the skin around the
be introduced. It is necessary to remove powerline interference
electrode. Motion artifacts resemble the signal characteristics
from ECG signals as it completely superimposes the low
of baseline wander, but are more problematic to combat since
frequency ECG waves like P wave and T wave. (Figure 3) shows
their spectral content considerably overlaps that of the PQRST
an ECG signal typically affected by a powerline interference.
complex. They occur mainly in the range from 1 to 10 Hz. In
the ECG, these artifacts are manifested as large-amplitude
waveforms which are sometimes mistaken for QRS complexes.

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101

 3
Electrode motion artifacts are particularly troublesome in the
context of ambulatory ECG monitoring where they constitute
π −ωc
the main source of falsely detected heartbeats. A typical ECG 1 1
∫ 1 ⋅ e jω n dω + ∫π 1⋅ e
jω n
h( n) = dω
signal affected by electrode motion artifact is shown in (Figure (2) 2π ωc 2π −
5) below.

 ωc
1− , n = 0
 π
=
− sin(ωc n) , n =±1, ±2...

 πn
(3)
truncation can be done by multiplying h(n) by a rectangular
window function, defined by

1, n = 0,...., L;
h( n) = 
(4) 0, otherwise
or by another window function if more appropriate. Such an
FIR filter should have an order 2L + 1 of approximately 1150 to
achieve a reasonable trade-off between stopband attenuation
Figure 5: ECG affected by electrode motion artifacts [2] (at least 20 dB) and the width of the transition band.

Matlab code to remove baseline wander using high-pass filter

2. Techniques to Remove Artifacts from ECG
clear all
Signal
Fs = 360; % Sampling Frequency

In this section, various signal processing methods for

N = 50; % Order
removing the artifacts from ECG signal have been described.
Fc = 0.667; % Cutoff Frequency
These methods are simple yet effective. The section also
includes the Matlab programs along with their results for the
% Construct an FDESIGN object and call its BUTTER method.
described methods.
h = fdesign. lowpass ('N,Fc', N, Fc, Fs);
Hd = butter(h);
2.1 Techniques for Removal of Baseline Wander

x=load('100.txt'); % load the ECG signal

A straightforward approach to the design of a filter is to
x1=x(:,2);
choose the ideal high-pass filter as a starting point [4],
x2=x1./ max(x1);

0, 0 < ω < ωc subplot (2,1,1), plot(x2), title ('ECG Signal with low-frequency
H (e jω ) =  (1) (baseline wander) noise'), grid on
1, ωc < ω < π
y0=filter (Hd, x2);
f
Where, 2π f c= 2π f c and
f
= . Thus, if fs
fs
fs
subplot (2,1,2), plot(y0), title ('ECG signal with baseline
= 0.5 Hz and f s =250 Hz then corresponding wander REMOVED'), grid on
normalized cut-off frequency ( f c ) = 0.002. Since the
corresponding impulse response has an infinite length,

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101

 4
d2 = wrcoef ('d', C, L,'bior3.7',2);
d1 = wrcoef ('d', C, L,'bior3.7',1);

y= d9+d8+d7+d6+d5+d4+d3+d2+d1;

subplot (2,1,2), plot(y), title ('ECG Signal after baseline wander

REMOVED'), grid on

(Figure 6): ECG signal with baseline wander (above); ECG

signal with baseline wander removed (below) using high-pass
filter.

Wavelet transform can also be used to remove the baseline

wander from ECG signal. The frequency of baseline wander is
approximately 0.5 Hz. According to discrete wavelet transform
(DWT), the original signal is to be decomposed using the
subsequent low-pass filters (LPF) and high-pass filters (HPF).
The cut-off frequency for LPF and HPF will be half of the
sampling frequency. For example, if the sampling frequency
is 250 Hz, then 125 Hz will be the cut-off frequency for both Figure 7: ECG signal with baseline wander (above); ECG signal
LPF and HPF in the first level decomposition. In second level with baseline wander removed (below) using DWT.
decomposition, the cut-off frequency becomes 62.5 Hz, for
2.2 Techniques for Removal of Powerline Interference
third level decomposition it becomes 31.25 Hz and so on.
Thus, it will require nine- level decomposition using DWT to A very simple approach to the reduction of powerline
remove a baseline wander of 0.5 Hz frequency. Following is the interference is to consider a filter defined by a complex-
Matlab code to remove baseline wander from an ECG signal conjugated pair of zeros that lie on the unit circle at the
using DWT. interfering frequency [4],
z1,2 = e ± jω0

Matlab code to remove baseline wander using DWT z1,2 = e ± jω0 (5)
x=load ('100.txt');
x1=x (:,2); Such a second-order FIR filter has the transfer function
x2=x1. /1000;
x2=x2 (170000:215000);
(1 − z1 z −1 )(1 − z2 z −1 )
H ( z) =
subplot (2,1,1), plot(x2), title ('ECG Signal with baseline (6)
wander'), grid on 1 − 2cos(ω0 ) z −1 + z −2
=
[C, L] = wavedec (x2,9,'bior3.7'); % Decomposition
Since this filter has a notch with a relatively large bandwidth,
a9 = wrcoef ('a', C, L,'bior3.7',9); % Approximate Component
it will attenuate not only the powerline frequency but also the
d9 = wrcoef ('d', C, L,'bior3.7',9); % Detailed components
d8 = wrcoef ('d', C, L,'bior3.7',8); ECG waveforms with frequencies close to ω0 . It is, therefore,
necessary to modify the filter in (6) so that the notch becomes
d7 = wrcoef ('d', C, L,'bior3.7',7); more selective, for example, by introducing a pair of complex-
d6 = wrcoef ('d', C, L,'bior3.7',6);
d5 = wrcoef ('d', C, L,'bior3.7',5); conjugated poles positioned at the same angle as the zeros p1,2 = re ± jω0
but at a radius r,
d4 = wrcoef ('d', C, L,'bior3.7',4);
d3 = wrcoef ('d', C, L,'bior3.7',3); p1,2 = re ± jω0 (7)

JScholar Publishers
J Biomed Eng Res 2019 | Vol 3: 101
 5

Where 0 < r < 1 . Thus, the transfer function of the

resulting IIR filter is given by

(1 − z1 z −1 )(1 − z2 z −1 )
H ( z) =
(1 − p1 z −1 )(1 − p2 z −1 )
(8)
1 − 2cos(ω0 ) z −1 + z −2
=
1 − 2r cos(ω0 ) z −1 + r 2 z −2

The notch bandwidth is determined by the pole radius r and

is reduced as r approaches the unit circle. Figure 8 shows
the impulse response and the magnitude function for two
different values of the radius, r = 0.75 and 0.95. From Figure
8 it is obvious that the bandwidth decreases at the expense of
increased transient response time of the filter. The practical
implication of this observation is that a transient present
in the signal causes a ringing artifact in the output signal.
For causal filtering, such filter ringing will occur after the
transient, thus mimicking the low-amplitude cardiac activity
that sometimes occurs in the terminal part of the QRS
complex, i.e., late potentials [4].

Figure 8: Pole-zero diagram for two second-order IIR filters

whose zeros are identically positioned but whose poles are at a
radius r of either 0.75 or 0.95. The impulse response h (k) and
the corresponding magnitude function are shown in the left
and right panels, respectively [4].

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101

 6
Matlab code to remove powerline interference from
ECG signal
Clear all

Fs = 360; % Sampling Frequency

Fnotch = 0.67; % Notch Frequency
BW = 5; % Bandwidth
Apass = 1; % Bandwidth Attenuation

[b, a] = iirnotch (Fnotch/ (Fs/2), BW/(Fs/2), Apass);

Hd = dfilt.df2 (b, a);

x=load ('100.txt');
x1=x (:, 2);
x2=x1. / max(x1);
Subplot (3, 1, 1), plot(x2), title ('ECG Signal with baswline
wander'), grid on (Figure 9): Original ECG signal containing both baseline
wander and powerline interference (top); ECG signal with
y0=filter (Hd, x2); baseline wander removed (middle); ECG signal with powerline
Subplot (3, 1, 2), plot(y0), title ('ECG signal with low-frequency interference removed (bottom)
noise (baswline wander) Removed'), grid on
2.3 Techniques for Removal of Electromyographic (EMG)
Noise
Fnotch = 60; % Notch Frequency
BW = 120; % Bandwidth
The EMG noise is a high-frequency noise; hence an n-point
Apass = 1; % Bandwidth Attenuation
moving average (MA) filter may be used to remove, or at least
suppress, the EMG noise from ECG signals. The general form
[b, a] = iirnotch (Fnotch/ (Fs/2), BW/ (Fs/2), Apass);
of an MA filter is
Hd1 = dfilt.df2 (b, a);

n
y1=filter (Hd1, y0);
Subplot (3, 1, 3), plot (y1), title ('ECG signal with power line
=y ( n) ∑
k =0
bk x(n − k ) (10)
noise Removed'), grid on
Where x and y are the input and output of the filter, respectively.
The bk values are the filter coefficients or tap weights, k = 0,
The above Matlab code implements two IIR notch filters: one 1, 2, . . . , N, where N is the order of the filter. The effect of
for removing the baseline wander with a notch concentrated at division by the number of samples used (N + 1) is included in
0.67 Hz and another for removing the powerline interference the values of the filter coefficients. The signal-flow diagram of
a generic MA filter is shown in (Figure 10) [5].
with a notch concentrated at 60 Hz. The results of the code are
shown in (figure 9) below.

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101

 7
in the result. This is due to the fact that the attenuation of
the simple 8-point MA filter is not more than -20 dB at most
frequencies (except near the zeros of the filter) [5].

(Figure 10): Signal-flow diagram of a moving-average filter of

−1
order N. Each block with the symbol z represents a delay of
one sample, and serves as a memory unit for the corresponding
signal sample value [5].

Increased smoothing may be achieved by averaging signal

samples over longer time windows, at the expense of increased
filter delay. If the signal samples over a window of eight samples
are averaged, we get the output as [5]

1 7 −k
H ( z ) = ∑ z (11)
8 k =0 (Figure 11): ECG signal with high-frequency (EMG like)
noise; fs = 1,000 Hz [5]
The transfer function of the filter is

1 7 −k
H ( z ) = ∑ z (12)
8 k =0
The 8-point MA filter may be rewritten as

1 1
y (n)= y (n − 1) + x(n) − x(n − 8)
8 8 (13)

The recursive form as above clearly depicts the integration

aspect of the filter. The transfer function of this expression is
easily derived to be

(Figure 12): The ECG signal with high-frequency (EMG like)

1 1 − z −8  noise in (Figure 11) after filtering by the 8-point MA filter [5]
H ( z) =  (14)
8 1 − z −1  Another approach to dealing with this problem is offered by
time-varying lowpass filtering using a filter with a variable
(Figure 11) shows a segment of an ECG signal with high-
frequency response. For example, a filter with a Gaussian
frequency noise. (Figure 12) shows the result of filtering the impulse response has been suggested for this purpose as the
signal with the 8-point MA filter described above. Although filters bandwidth is easily changed from one sample to another
the noise level has been reduced, some noise is still present through a function β (n) which defines the width of the

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101

 8
Gaussian [4], Matlab code to remove (electrode) motion artifacts
h( k , n) ~ e − β ( n ) k
2
(15)
from ECG

The width function β ( n) is designed to reflect local signal Clear all

properties so that smooth segments of the ECG are subjected y1=load ('ECG1.txt'); % this is an ECG signal with motion
to considerable low-pass filtering, whereas the QRS interval, artifacts
with its much steeper slopes, largely remains unfiltered. By
making β ( n) proportional to the derivative of the signal, y2= (y1 (:,1)); % ECG signal data
slow signal changes produce small values of β ( n) , thus a1= (y1 (:,1)); % accelerometer x-axis data
making the Gaussian impulse response to decay more slowly to a2= (y1 (:,1)); % accelerometer y-axis data
zero so as to produce greater noise suppression, and vice versa. a3= (y1 (:,1)); % accelerometer z-axis data
Details of designing the width function β ( n) , truncating h
(k, n) in (15), and the resulting performance on ECG signals y2 = y2/max (y2);
can be found in [5]. The idea of adapting the cut-off frequency
of a linear low-pass filter to the slopes of the ECG has also Subplot (3, 1, 1), plot (y2), title ('ECG Signal with motion
been explored for other types of filters [4, 6 – 8]. For more artifacts'), grid on
mathematical details of time-varying filters, references [6 – 9]
may be referred. a = a1+a2+a3;
a=a/max (a);
2.4 Techniques for Removal of Electrode Motion
Artifacts
Mu= 0.0008;
One of the widely used techniques for removing the electrode Hd = adapt filt. Lms (32, mu);
motion artifacts is based on adaptive filters. The general [s2, e] = filter (Hd, a, y2);
structure of an adaptive filter for noise canceling utilized in
this paper requires two inputs, called the primary and the Subplot (3, 1, 2); plot (s2), title ('Noise (motion artifact)
reference signal. The former is the d(t) = s(t) + n1(t) where s(t) estimate'), grid on
is an ECG signal and n1(t) is an additive noise. The noise and
the signal are assumed to be uncorrelated. The second input is Subplot (3, 1, 3), plot (e), title ('Adaptively filtered/ Noise free
a noise u(t) correlated in some way with n1(t) but coming from
another source. The adaptive filter coefficients wk are updated ECG signal'), grid on
as new samples of the input signals are acquired. The learning
rule for coefficients modification is based on minimization, in
the mean square sense, of the error signal e(t) = d(t) − y(t)
where y(t) is the output of the adaptive filter. A block diagram
of the general structure of noise cancelling adaptive filtering
is shown in figure 13 [10]. The two most widely used adaptive
filtering algorithms are the Least Mean Square (LMS) and the
Recursive Least Square (RLS).

Figure 14: Removal of motion artifacts from ECG using

adaptive filtering. (top) ECG signal with motion artifacts,
(middle) noise/ motion artifact, (bottom) noise-free ECG
Figure 13: Block diagram of adaptive filtering scheme [10] signal

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101

 9
References
1. Available from: https://www.researchgate.net/publica-
tion/302631407_Low-Power_Wearable_ECG_Monitor-
ing_System_for_Multiple-Patient_Remote_Monitoring/
figures?lo=1

2. Available from: https://www.researchgate.net/publica-

tion/224326734_Estimation_of_noise_in_ECG_signals_us-
ing_wavelets/figures?lo=1

3. A. Jayant, T. Singh and M. Kaur (2013): Different Tech-

niques to Remove Baseline Wander from ECG Signal, Int. J. of
Emerging Research in Management & Technology, 2.

4. Leif Sornmo and Pablo Laguna. (2005) Bioelectric Signal

Processing in Cardiac and Neurological Processing. 1st Ed.,
Elsevier Academic Press, ISBN: 9780124375529.

5. Rangraj M. Rangayyan. (2002) Biomedical Signal Analysis:

A case study approach. John Wiley & Sons, Inc., ISBN: 0-471-
20811-6.

6. J. L. Talmon, J. A. Kors, and J. H. van Bemmel (1986): Adap-

tive Gaussian filtering in routine ECG/VCG analysis, IEEE
Trans. Acoust. Speech Sig. Proc. 34: 527-534.

7. E. K. Hodson, D. R. Thayer, and C. Frankli (1981): Adaptive

Gaussian filtering and local frequency estimates using local
curvature analysis, IEEE Trans. A coust. Speech Sig. Proc. 29,
854-859.

8. V. de Pinto (1991): Filters for the reduction of baseline wan-

der and muscle artifact in the ECG, J. Electro cardiol. 25: 40-
48.

9. Roman Kaszynski and Jacek Piskorowski (2007): Selected

Structures of Filters with Time-Varying Parameters, IEEE
Transactions on Instrumentation and Measurement, 6: 2338-
2345.

10. M. Milanesi, N. Martini, N. Vanello, V. Positano, M. F. Submit your manuscript to a JScholar journal
Santarelli, R. Paradiso, D. De Rossi and L. Landini. (2006) and benefit from:
Multichannel Techniques for Motion Artifacts Removal from ¶¶ Convenient online submission
Electrocardiographic Signals. In: Proceedings of the 28th IEEE
¶¶ Rigorous peer review
EMBS Annual International Conference New York City, USA.
3391-3394 ¶¶ Immediate publication on acceptance
¶¶ Open access: articles freely available online
¶¶ High visibility within the field
¶¶ Better discount for your subsequent articles
Submit your manuscript at
http://www.jscholaronline.org/submit-manuscript.php

JScholar Publishers J Biomed Eng Res 2019 | Vol 3: 101