This full-text paper was peer-reviewed and accepted to be presented at the IEEE ICCSP 2015 conference.

Removal of Noise from Electrocardiogram Using

Digital FIR and IIR Filters with Various Methods
K.Sravan Kumar, Babak Yazdanpanah, P Rajesh Kumar

Abstract-Electrocardiogram (ECG) is a type of measuring An ECG is very sensItIve, different types of artifact and
the electrical activities of heart. Each section of ECG is necessary interference can contaminate the ECG signal as the real
for the diagnosis of various cardiac problems. But the amplitude
amplitude and time period of the signal can be changed. ECG
and time period of ECG signal is generally corrupted by various
noises. Mter an analog ECG signal is transformed into digital
signals are mostly affected by:
format, appropriate digital filter can be utilized to repress the
1) Baseline Wander: It is actually the effect where the base
various kinds of noise like Baseline Wander, Power line
Interference, High -frequency Noise, Physiological Artifacts etc.,
axis of any ECG signal viewed on a screen appears to
depends on their specifications. In generic two types of method "wander" or move up and down rather than be straight. This
can be classified in this paper; FIR filters like Rectangular, cause the entire signal to shift from its normal base. It is
Hann, Blackman, Hamming and Kaiser window techniques and usually caused due to improper electrodes and by respiration
IIR filters like Butterworth, Chebyshev I, Chebyshev II and
or movement of the patient.
Elliptic filters are also prospected to reduce artifacts in ECG
signal. The results are collected from different orders for FIR 2) Physiological Artifacts: This kind of noises is primarily
filter as 56, 300, 450, and 600 and for IIR filter as 1, 2, and 3. The
produced by other organs of the body or caused by muscular
signals taken from the MIT-BIH data base which contains the
normal and abnormal waveforms. The work has been
retractions relevant to breathing. Electrode movement artifact
implemented in MATLAB FDA Tool. The results are obtained is commonly considered to be the most difficult, since it may
using different window based FIR filters, IIR filter with different imitate the figure of abnormal beats and cannot be eliminated
approximation methods and their respective waveforms are simply by normal filters, as can an artifact of other kinds.
shown. In addition, power spectrum density, signal to noise ratio
(SNR) and means square error (MSE) of both noisy and filtered 3) High -frequency Noise: This kind of random artifact
ECG signals are calculated. We observed that Digital FIR filter could be due to the thermal cause in electrodes, the recording
with Kaiser Window in order 56 shows high performance as
system, , the instrumentation amplifiers and pickup of ambient
compared to the other windowing techniques and Digital IIR
filter approximation methods.
electromagnetic signals by cables[ I].In the real-time clinical
monitoring systems in surgery, electrosurgical nOIse IS a
Index Terms-FIR filters, IIR filter, Electrocardiogram, SNR, considerable barrier to be overcome.
MSE,MATLAB
4) Power line Interference: Interference caused by radiation
LINTRODUCTION from the high voltage power for Examples of environmental
interference are those effected by 50 or 60 Hz power-supply
Signal processing has a wealthy background and its
lines, , radiation from lights, electrodes motion and radio­
significance in biomedical engineering is known to all. An
frequency emissions from nearby medical devices.
electrocardiogram demonstrates the electrical activity in the
heart, and may be analyzed in characteristic parts, named P, Q, These artifacts and interference makes the wrong diagnosis
R, S, and T waves. When an ECG is recorded, it would be of the ECG signal [1-2].So, the elimination of these artifacts
corrupted with various kinds of noise. Therefore exploitation and interference from ECG signal has become very difficult.
of pure ECG indices from noisy measurements has been one Different kinds of digital filters have been used to rectify the
of the important considerations of biomedical signal problem [2-4].But it is hard to utilize these filters with fixed
processing and requires trusty techniques to maintain the coefficients to remove several kinds of artifacts, because the
diagnostic data of the recorded signal. ECG signal is known as a non-stationary signals.

A number of researchers have worked on the digital filters

for eliminating from ECG signal and various techniques have
been offered or proposed. Chandrakar et al. studied FIR and
K.Sravan Kumar is with the Center for biomedical engineering, Department IIR filters then concluded that Kaiser Window based FIR
of Electronics and Communication Engineering, visakhapatnam, 530003 filters are better to remove artifacts from ECG signals
India (phone: +91-9701287223; email:[email protected]).
Babak Yazdanpanah is with the is with the Center for biomedical
[5].They implemented efficient hardware architecture for high
engineering, Department of Electronics and Communication Engineering, speed FIR filter for electromyogram removal from
Visakhapatnam, 530003 India (e-mail: [email protected]). electrocardiogram signals [6]. FIR digital filters were designed
Dr. P. Rajesh Kumar is with the Department of Electronics and with new window to remove the interferences or the artifacts
Communication Engineering, Visakhapatnam, 530003 India (Phone: +91-
9440367694, email: [email protected]).
and calculated SNR of noiseless ECG compared with existing

978-1-4 799-8081-9/15/$3l.00 © 2015 IEEE

0157
 This full-text paper was peer-reviewed and accepted to be presented at the IEEE ICCSP 2015 conference.

windows [7]. They proposed second order infinite impulse the frequency response of the IIR being convolved with the
response (IIR) notch filter, adaptive notch filtering technique Fourier transform of the window function. If the window's
with least mean square algorithm and discrete wavelet main lobe is narrow, the composite frequency response
transform method for the removal of power line interference remains close to that of the ideal IIR filter.
from ECG signal and considered two synthesis parameters
MSE and SNR [8]. Mohandas Choudhary, Ravindra Pratap III. METHODOLOGY
Narwaria "Suppression of Noise in ECG Signal Using Low ECG signal fundamentally contains frequency between 0-
pass IIR Filters" describes an IIR filter design for noise 250Hz. Research proves that the frequency range of the ECG
removing [9] it gives various types of IIR filter design but the signal is 0-250 Hz, The sampling frequency was selected to
result of Butterworth is best among all. Mbachu et al. designed facilitate performances of 60 Hz digital notch filter in
digital FIR low pass, high pass and notch filters with a arrhythmia detectors, sampling frequency of data signal is 360
rectangular window for the removal of the artifacts [10]. In Hz and amplitude Imv. We designed the filter for corrupted
this work, the artifact from the electrocardiogram waveform ECG signal in four steps: In the first, step with the help of
has been eliminated utilizing these various methods. Signal to FDA Tool in MATLAB software design digital FIR filters
noise ratio (SNR), Means square error (MSE) and power with various windowing techniques like: Rectangular, Hann,
spectral density resulting from the algorithms has been Hamming, Blackman and Kaiser with high pass filter cut off
investigated and compared. frequency 0.5 Hz were utilized to delete baseline wander
noise from noisy ECG signal. In the second step, removing
The rest of this article organized as follows: Section II
theoretical description of digital filters. Section III summarizes power line interference (50/60 Hz) by band stop with cut off
the techniques investigated. Section IV provides the results frequency (59.5Hz-60.5 Hz) was carried out , in the third step,
and output waveforms. Finally, some conclusions are we deleted EMG noise by applying low pass filter with cut off
frequency 100Hz, in order 56,300,450 and 600, finally moving
demonstrated in Section V.
average filter to smooth the ECG waveform. The task was
II.THEORY accomplished in various orders.

A. DigitalllRfilter HPF BSF lPF

Digital Filters are designed by using the values of both the

I NOI� EC) (High pass
filter)
I--- (Band stop
filter)
t-- (low pass
filter)
past outputs and the present input, an action brought about by
convolution. If such a filter is subjected to an impulse then its
Fig. 1. Block diagram ofECG signal processor
output not requirement essentially become zero. The impulse
response of such a filter may be infinite in period. Such a filter In addition, we designed IIR digital filter with different
is known as an Infinite Impulse Response filter or IIR filter. approximation methods like Butterworth, Chebyshev I,
The infinite impulse response of such a filter implies the Chebyshev II, and Elliptic in order 1, 2 and 3 with the low
capability of the filter to have an infinite impulse response.
pass, high pass, and band stop filters the specifications
This indicates that the system is prone to feedback and
mentioned above. The efficiency analysis contained the
inconstancy. IIR filters are designed fundamentally by the
comparison of outcomes generated by filters designed during
Impulse immutability or the Bilinear Transformation method.
the modelling method by replacing filter parameters arranging
The IIR filter is determined by the equation below:
them so that the one where outcomes obtained were best.
(1)
The results were collected with the performing filters
B. Digital FIRfilter through the ECG database from MIT-BIH site. The ECG
samples 113m, 114m, 116m, 118m, 124m, 201m, 203m,205m
FIR filters are digital filters with finite impulse response. (MLII, VI) obtained from MIT have supported then own
They are also known as non-recursive digital filters as they do research into arrhythmia database which consists 48 half hours
not have the feedback, even though recursive algorithms can excerpts of two channel ambulatory ECG recordings utilized
be used for FIR filter realization. FIR filter is characterized by to verify the results of digital filter designed as described in
following equations:
the above methodology.
o ·
Id'
y(m) = i. Q"y (m - k) - L b"x (m - k) (2)
k.l ".0
IV. RESULTS
In the window design method, one first designs an ideal IIR
The results were generated with the designed FIR, IIR filters
filter and then truncates the infinite impulse response by
applied to various ECG signals from MIT-BIH database. The
multiplying it with a finite length window function. The result
filters like high pass , band stop and low pass with various
is a finite impulse response filter whose frequency response is
modified from that of the IIR filter. Multiplying the infinite windowing at order 56, 300, 450, 600 and IIR with different
impulse by the window function in the time domain results in approximation methods in order 1, 2 and 3 shows different
results.

0158
 This full-text paper was peer-reviewed and accepted to be presented at the IEEE ICCSP 2015 conference.

TABLE I TABLE III

SNR COMPARSION OF WINDOW BASED FIR FILTERS SNR COMPARSION OF IIR APPROXIMATION METHOD
Real SNR Signal to Noise Ratio of the IIR filtered ECG signal
ECG of
Real SNR Signal to Noise Ratio of FIR filtered ECG signal Data noisy Butterworth Chebyshev Chebyshev
ECG of Elliptic
ECG
Data noisy Recta- Hann Ham- Blackman Kaiser Type I Type 2
ECG ngular ming
113 5.77 5.70 5.37 1.61 5.37
113 5.77 6.72 6.13 6.17 6.05 6.69
114 1.36 1.94 1.39 2.62 1.39
114 1.36 2.52 1.82 1.87 1.72 2.49
116 8.00 8.40 8.52 0.94 8.52
116 8.00 9.13 9.02 9.03 8.91 9.16
118 8.96 8.64 8.42 10.35 8.42
118 8.96 10.02 9.29 9.35 9.20 10.05
124 8.62 9.46 8.60 0.74 8.60
124 8.62 9.86 9.03 9.09 8.92 9.89
201 14.37 14.28 14.33 8.16 14.33
201 14.37 14.47 14.53 14.53 14.51 14.48

203 12.84 13.71 13.08 13.13 13.01 13.76 203 12.84 12.31 12.61 3.10 12.61

205 9.02 9.94 9.50 9.54 9.41 9.95 205 9.02 9.02 9.01 5.47 9.01

TABLE II TABLE IV
MSE COMP ARSION OF WINDOW BASED FIR FILTERS MSE COMPARSION OF IIR APPROXIMATION METHOD

Real MSE MSE of IIR filtered ECG signal

ECG of
Real MSE Mean Square Error of the FIR filtered ECG signal data noisy Butterworth Chebyshev Chebyshev Elliptic
ECG of ECG
Data noisy Rect- Hann Hamm- Blackman Kaiser Type I Type 2
ECG angular ing
113 0.21 0.14 0.15 2.59 0.15
113 0.21 0.17 0.19 0.19 0.18 0.17
114 0.03 0.01 0.11 3.48 0.01
114 0.03 0.02 0.03 0.03 0.03 0.01
116 1.26 0.34 0.35 5.55 0.35
116 1.26 0.98 0.10 1.09 1.12 0.97
118 1.09 0.15 0.16 6.27 0.16
118 1.09 0.81 0.94 0.93 0.96 0.80
124 0.76 0.11 0.12 4.20 0.12
124 0.76 0.56 0.65 0.64 0.67 0.55
201 0.06 0.14 0.14 2.56 0.03
201 0.06 0.05 0.62 0.06 0.06 0.05

203 0.29 0.24 0.25 0.25 0.02 0.23 203 0.29 0.03 0.03 3.87 0.21

205 0.15 0.11 0.13 0.13 0.13 0.10 205 0.15 0.19 0.21 6.80 0.03

The graphs for the waveforms and their power spectral

density before and after filtering are shown for various
windows and IIR filters.
By using these windows and IIR approximation methods, we
designed the high pass filter of cut-off frequency 0.5 Hz for
removing baseline wandering, band stop filter of cut-off
frequency 59.50Hz - 60.5 Hz for removing power line
interference and low pass filter of cut-off frequency 100Hz for
removing EMG noise. The comparison of different FIR and
IIR filters by calculation of SNR (Signal to noise ratio) and
MSE (mean square error) was done at 56 and 1 orders. The
results are shown in tabular form.

Fig. 2. Noisy ECG (113m) and removed BLW using Kaiser Window

0159
 This full-text paper was peer-reviewed and accepted to be presented at the IEEE ICCSP 2015 conference.

: ;�UIDI�I II� 1
ECG Si n l with EMG Noise
ga
ECG Signal without Powerline I nterference
3000 r---�----�--r---�---.

g 2000

o 10 20 30 40 50 60 70 80 90 c..

Time I s Ji 1000

:;� j o 10 20 30 40
Time I s
50 60 70 80 90 3ooo r----,-----.--.---�
ECG Si gan
Frequency I Hz
l wth reduced EMG Noise

� 2000

:;� l
"
c.
Ji 1000
o 10 20 30 40 50 60 70 80 90
Time I s
....&..
30�0--- --
1�
5 0---- 1 �--�
-�00 -50--�1I ..�5� 1� �--�
0 ----00 1 5�
0 ---)
200
Frequency I Hz
Fig. 3. Remove PLI (60 Hz) and EMG using Kaiser Window
Fig. 6. Frequency spectrum of with and without EMG using Kaiser Window

E C G Signal with Baseline Wander Periodogram Power Spectral Density Estimate

6000 20 ,----,----.---.--,

I , , I , , •
Q)
'" 4000 o - - - - . L. ______ ... . ______ .. _______ � _______ .. _______.. ______ -'. ______ .. ______
�
I , I I , I I ,
I , I I , I I •
, . , , . , , .
0.
� 2000
N
-
20 - - - - - - - � - - - - - - - r - -------
, .
: -------:- - - - - - - i - -----
, , .
:r:
rn
0 � I , , • ,

-200 -150 -100 -50 0 50 100 150 200 >-. 40 I

,
I I
..
I

,
I
- - - - - - - � - - - - - - - � - ------ ------ .,- - - - - - - � - -----

u , ,
co , ,
Frequency I Hz Q)
�
E CG Signal with reduced Baseline Wander <T
i" �O
3000
�
�
CL
Q)
2000 �o
'"
�
li
� 1000 -
100

..oJ iii..
0 -
120
-200 -150 -100 -50 0 50 100 150 200 0 20 40 60 80 100 120 140 160 180
Frequency J Hz Frequency (Hz)

Fig. 4. Frequency spectrum of with and without BL W using Kaiser Window Fig. 7. Power spectrum density of noisy ECG signal using Kaiser Window

ECG Signal with Powerline Interference Periodogram Power Spectral Density Estimate
3000 20

., 2000 0 -'
______I.... _____ _______ .>_______ L _______ L _______I...______ . _____ • .>. _____ -'
I I I I I I I I
I I I I I I I I
I I I I I I I I
� I I I I I I I I
c. I I I I I I I
I I I I I I I
E I I I I I I I I

<! 1000 _____ .J _______ J _______ L _______ L _______ ______ .J _______ J ______
I I I I I I I
'-
N I I I I I I I
I I I I I I I

...&.. �� I
"
I

,
,
I I
,
,
,
0 '--" -- r ' --" -- r ' --" --" i r ]--" - ---" - '--" -

-200 -150 - 01 0 -50 0 50 100 51 0 200 >-.

u
co
Frequency I Hz .,
�
<T
ECG Signal ""th reweed Powenine Interference , , ,
i" �o -------r-------"'l-------,------

3000 'S ,
,
,
,
,
,
.
.
.
, , .
�
CL
,
,
,
,
,
,
,
,
,
�o -----r------,-------,------

� 2000
.�
Ci
100
- .. _______ � _______ .. ------
Ji
- - - - - -I- ______ -0 _______ ------1--------4-------
, , .
1000 I
I
I
I
I
I
I
I
I
I
I
I
,
,
,
,
.
,
I I I I , , ,
I I I I , , ,
, , ,
...&...
I I I I

-120 "----'----"---'---"---'
0 o 20 40 60 80 100 120 140 160 180
-
200 -
150 -100 -50 0 50 100 150 200 Freq u ency (Hz)
Frequency I Hz

Fig. 5. Frequency spectrum of with and without PLI using Kaiser Window Fig. 8. Power spectrum density of denoised ECG signal using Kaiser Window

0160
 This full-text paper was peer-reviewed and accepted to be presented at the IEEE ICCSP 2015 conference.

EGG Signal with Baseline Wander

V. CONCLUSION 6000

This article, introduced a method used for artifacts reduction

'" 4000
from ECG signal which basically includes designing of filter =

'"
�

0.
with different windows like Rectangular, Hann, Hamming, E
<{ 2000
Blackman and Kaiser Window with different order
56,300,450,600 and IIR filters approximation methods like 0
.,111.
-200 -150 -100 -50 0 50 100 150 200
Butterworth, Chebyshev I, Chebyshev II and Elliptic at orders Frequency I Hz
EGG Signal wlh reduced Baseline Wander
1, 2, 3 with proper parameters indicating the best outcomes of 3000
baseline wander noise, power line interference and
electromyography noise removal. The results for various '" 2000
=
�

filters are considered and evaluated by waveforms, power 0.

E
spectrums density (PSD), signal to noise ratio (SNR) and <{ 100 0

Mean Square Error (MSE).The Kaiser Window order 56 of

0
� La..
FIR filter showing the best results was compared to all orders -200 -150 -100 -50 0 50 100 150 200
Frequency I Hz
in FIR as well as IIR filters at order 1, 2 and 3.
Noisy E CG Signal Fig. 11. Frequency spectrum of with and without BL W using
4 Butterworth approximation method

> EGG Signal with Powerline Interference

E 2 3000 r----r----�--__,
-
'"
F '"
(5 0 =
2000
>
�

0.
E
<{ 1000
-2
0 10 20 30 40 50 60 70 80 00
lime/s .... �.
0
ECG Signal without Basel ine Wander -200 -150 -100 -50 50 100 150 200
3 Frequency I Hz
EGG Signal with reduced Powerline Interference

> 2
3000

E
'" '"
-0
2000
� �

0.
> 0
0
E
<{ 1000

-1
0 10 20 30 40 50 60 70 80 00 •• �.
0
lime/s -200 -150 -100 -50 50 100 150 200
Frequercy I Hz

Fig. 9. Noisy ECG (l13m) and removed BLW using Butterworth Fig. 12. Frequency spectrum of with and without PLI using
approximation method Butterworth approximation method

EGG Signal without Powe�ine Interference

I�� . o 10 20 30 40
Time / s
50 60 70 80 90

EGG Signal without EM G Noise

i:� �
>
.
0 10 20 30 40
Time / s
50 60 70 80 90

Smo othed EGG Signal

��� .
-> 0 10 20 30 40
Time / s
50 60 70 80 90

Fig. 10. Remove PLI (60 Hz) and EMG using Butterworth Fig. 13. Frequency spectrum of with and without EMG using
approximation method Butterworth approximation method

0161
 This full-text paper was peer-reviewed and accepted to be presented at the IEEE ICCSP 2015 conference.

REFERENCES

[I] E. T. Gar, C. Thomas and M. Friesen, "Comparison of Noise

Sensitivity of QRS Detection Algorithms;' IEEE Tran. Biomed. Eng.,
vol. 37, no.I, pp. 85-98, January 1990.
[2] M. Kaur and B. Singh,"Power Line Interference Reduction in ECG
Using Combination of MA Method and llR Notch Filter;' Int. J. Of
Recent Trends in Eng., vol. 2, no. 6, pp. 125-129, November 2009.
[3] Y. Kumar and G. K. Malik, "Performance Analysis of different Filters
for Power Line Interface Reduction in ECG Signal;' Int. J. of Computer
Applications (0975 - 8887), vol. 3, no.7, pp. 1-6, June 20IO.
[4] M. S. Chavan, R. Agarwala, M. D. Uplane, and M. S. Gaikwad,"Design
of ECG Instrumentation and Implementation of Digital Filter for Noise
Reduction;' World Scientific and Engineering Academy and Society
(WSEAS). Stevens Point, Wisconsin, USA. vol. I. no. 157-474. pp. 47-
50. January 2004.
[5] B.Chandrakar.O.P.Yadav and V.K.Chandra,"A Survey Of Noise
Removal Techniques For Ecg Signals ." [n [nternational Journal of A d
vanced Research in Computer and Communication Engineering vol. 2
• Issue 3 pp.1354-I357.March 2013.
•

[6] M.Leelakrishna. And ISelvakumar, "Fpga Implementation of High

Fig. 14. Smoothed ECG (113) Signal (FFT) for Butterworth Speed Fir Low Pass Filter for Emg Removal from ECG.'"in International
approximation method Journal of Engineering Research & Technology vol. 2 Issue 5, pp.855-
863.May 2013.
[7] D.V.L.N Sastry,K.S.Gupta ,V M.Krushna and D.S.y'S N. Reddy .
"Improved Snr Of Ecg Signal with New Window- Fir Digital Filters." in
International Journal of Advanced Research in Electronics and
Communication Engineering, vol I, Issue 3,pp.88-85 .September 20 12.
[8] S.B.S.Bagha."Removal Of 50hz Powerline Interference for Quality
Diagnosis of ECG Signal." in International Journal of Engineering
Science and Technology, vol.5, pp.1149- I I 55. May 2013.
[9] Mohandas Choudhary, Ravindra Pratap Narwaria"Suppression of Noise
in ECG Signal Using Low pass IIR Filters'", IJECSE 2012. ISSN- 2277-
1956.
[10] M.C.B. I.Victor, I.E.and N.LI , " Filtration Of Artifacts In ECG Signal
Using Rectangular Window-Based Digital Filters ." in International
Journal of Computer Science Issues, vol. 8. Issue 5. No.1. pp.279-
285.September 20 II.
[11] Sande Seema Bhogeshwar, M.K.Soni, Dipali Bansal.'"Design of
Simulink Model to denoise ECG signal using various IIR & FIR filters" .
2014 International Conference on Reliability, Optimization and
Information Technology -ICROIT 2014. India. Feb 6-8 2014.
[12] Thomas W. Parks, James H. Mc.Cleelan, "Chebyshev Approximation
for Nonrecursive Digital Filters with Filters with Linear Phase" . IEEE
Transactions on Circuit Theory, vol: 19, Issue: 2. pp.189-194,
Mar.l 972.
Fig. 15. Power spectrum density of noisy ECG signal using Butterworth
[13] Nalini Singh. Jhansi, Shahanaz Ayub, J.P. Saini."Design of Digital IIR
approximation method
Filter for Noise Reduction in ECG Signal'", 5th International Conference
on Computational Intelligence and Communication Networks, 2013.
[14] Vikrant Bhateja. Shabana Urooj, Rishendra Verma, Rini Mehrotra"A
20 ,---,----.---,--,---,,-
Novel Approach for Suppression of Powerline Interference and Impulse
Noise in ECG Signals'", IEEE International Conference on Multimedia.
Signal Processing and Communication Technologies (IMPACT).
pp.103-107. 23-25 Nov. 2013.
[15] Abd-el Hamid M. Abed . Gerald D. Cain ," The Host Windowing
Technique for FIR Digital Filter Design'" IEEE Transactions on
Acoustics, Speech and Signal Processing, voI.ASSP-32, No.4, pp-683-
694, Aug. 1984.
[16] J. A. Van Alste, T. S. Schilder, "Removal of Base-Line Wander and
Power-Line Interference from the ECG by an Efficient FIR Filter with a
Reduced Number of Tap" , IEEE Transactions on Biomedical
Engineering, voI.BME-32, No.12, pp-I052-1060, Dec. 1985.

Fig. 16. Power spectrum density of denoised ECG signal using

Butterworth approximation method

0162