Int'l Journal of Computing, Communications & Instrumentation Engg. (IJCCIE) Vol.

1, Issue 1 (2014) ISSN 2349-1469 EISSN 2349-1477

Analysis of Various Filter Configurations

on Noise Reduction in ECG Waveform
Prashanth Shetty, and Somashekhara Bhat


Abstract— Electrocardiography (ECG) is the graphical recording For testing purpose, the ECG signals from MIT-BIH
of the electrical activity of the heart and plays a vital role in the Arrhythmia Database are taken as the test data [2]. All the
primary diagnosis and monitoring of the health of heart. The ECG work was done with MATLAB®. The noises were simulated
signal being very sensitive and weak in nature, highly prone to even
and added to the test data.
small noise. Hence the design of the entire efficient filter system is
the key for the success of the entire ECG processing system.
In this work our main task is to design efficient analog filters to
eliminate all the noise sources associated with the ECG signal so as
to get a noise free ECG signal as output from the filters. This paper
mainly focuses on filtering the noise sources mainly corrupt the
retrieved ECG signal such as power line interference, base line
wander noise, motion artifact noise instrumentation noise etc.
The main aim of this work is to study the effect of various filtering
stages on the noisy ECG signal, to approximately estimate the order
of each filter stage required to get the required quality of output
signal and to show the proof of concept of the entire filtering process.
Several types of analog filters were designed, implemented and tested
for their correctness.

Keywords— Analog Filters, Biomedical Signal Processing,

Butterworth and Chebyshev filter, Electrocardiogram, Elliptic filter,
Signal to Noise Ratio (SNR).

I. INTRODUCTION

E LECTROCARDIOGRAPHY (ECG) is a non-invasive

method for capturing and processing the electrical
signature of the heart via skin electrodes and is in very wide
Fig. 1 Typical healthy Electrocardiogram trace [3]

II. NOISES AND ARTIFACTS IN ECG WAVEFORM

use (an estimated 200 million ECGs taken each year) [1].
Raw ECG data contain some noise and artifact components
ECG applications are broad and varied in their scope, and
that alter the shape of the ECG trace from the ideal structure as
therefore, so are the requirements for the analog data
shown in Fig. 1 and render the clinical interpretation
acquisition and processing system. As with most of the
inaccurate and misleading; consequently, a pre-processing step
biomedical signals, the ECG signal and its harmonics are low for improving the signal quality is a necessity. It is therefore
in bandwidth (0.1 to 300 Hz) with typical amplitude of 0.1 to 4 important to be familiar with the most common types of noise
mV as shown in Fig. 1. For this reason, a high dynamic range and artifacts in the ECG and the type of filters required to
is required for this system. The increasing demand on portable denoise the ECG.
and wearable personal healthcare devices (e.g. heart rate ECG inevitably affected by various types of noise
monitoring watch) requires a new generation of Low Power interference, summed up in the following three types of noise
Low Voltage (LVLP) bio-medical signal processing building [4]:
blocks. The challenges in ECG signal acquisition revolve
A. Baseline Wander
primarily around external noise rejection on the analog front-
end (AFE) and finally signal processing by the back-end Baseline Wander is extraneous noise in the ECG trace that
conditioning circuitry. may be caused from a variety of noise sources including
perspiration, respiration, body movements, and poor electrode
contact. The magnitude of this wander may exceed the
Prashanth Shetty, Research Scholar, Department of Electronics and amplitude of the QRS complex by several times, but its
Communication, Manipal Institute of Technology, Manipal, INDIA– 576104; spectral content is usually confined to an interval below 1Hz
(e-mail: [email protected]).
Somashekhara Bhat, Professor, Department of Electronics and (between 0.15 and 0.3 Hz). Baseline drift can be represented
Communication, Manipal Institute of Technology, Manipal, INDIA – 576104. as a sinusoidal component at the frequency of respiration

http://dx.doi.org/10.15242/ IJCCIE.E1113078 88
 Int'l Journal of Computing, Communications & Instrumentation Engg. (IJCCIE) Vol. 1, Issue 1 (2014) ISSN 2349-1469 EISSN 2349-1477

added to the ECG signal. The drift causes problems in the noise. The performance of the filter was evaluated by
detection of ECG signals, e.g., sometimes the amplitude of T- comparing their power spectra along with the mean square
wave is higher than the peak of R-wave and results in false error (MSE).
detection of R-peak.
B. Simulation of Baseline Wander
B. Power Line Interference (50/60 HZ) Baseline wander is one of the most significant noise sources
Power Line Interference is high frequency noise caused by during the ECG measurement. This noise is simulated as a low
interferences from nearby devices as a result of improper frequency sinusoid. The frequency is 0.3 Hz and the amplitude
grounding of the ECG equipment. This noise degrades the is 1 mV for the simulation. It is modeled similar to the power
signal quality and affects the tiny features which can be critical line interference in (1).
for clinical diagnosis and monitoring and signal processing. The frequency content of baseline wander is usually in a
The power line interference is narrow-band noise centered at range well below 0.5 Hz (between 0.15 and 0.3 Hz). These
50 Hz (or 60 Hz) with a bandwidth of less than 1 Hz. low frequency components can severely affect the visual
interpretation of an ECG. In order to discard baseline drift, a
C. Electromyographic Noise (EMG Noise)
high pass filter designed with 0.5 Hz of cutoff frequency is
This is mainly caused by the electrical activity of skeletal required.
muscles during periods of contraction or due to a sudden body
movement. While the frequency component of EMG C. Simulation of Electromyographic Noise (EMG Noise)
considerably overlaps with that of the QRS complex, it also The frequency content of EMG signal is usually in a range
extends into higher frequencies. As a result, processing the from dc to 10 KHz. The high frequency components may be
ECG trace to remove this noise naturally results in introducing discarded using a low pass filter designed with 100 Hz of
some distortion to the signal. It has a frequency range between cutoff frequency. The Table I gives the filter type required for
dc and 10,000 Hz with amplitude of 10% level. removing the various sources of noise components.
There are other types of noises contaminating the ECG
signal such as instrumentation noise, electrosurgical noise and TABLE I
CONSOLIDATED FILTER TYPE FOR DIFFERENT NOISE REMOVAL
other less significant source of noise. Among these noises, the
power line interference and the baseline wandering are the Frequency
S. No Noise Type Filter Used
Range
most significant and can strongly affect ECG signal analysis.
Except for these two noises, other noises may be wideband and 1 Baseline Wander High Pass Filter Below 1 Hz
usually a complex stochastic process which also distort the
Power Line
ECG signal. 2 Band Stop Filter 50/60 Hz
Interference

III. SIMULATION AND FILTERING OF NOISE IN ECG 3 EMG Noise Low Pass Filter Above 100 Hz
Since the goal of this paper is to design and implement
analog filters for the filtering of noises and to evaluate the
A time domain plot of noise free ECG signal from database
performance of these filters, each noise signal is first modeled
and the ECG contaminated by power line interference and
and simulated. The characteristics of each noise described in
baseline wander is described in Fig. 2, 3.
section 2 play a vital role in modeling the noise signal.
A. Simulation of Power Line Interference [5]
Power line interference consists of 50/60 Hz and its
harmonics which can be modeled as sinusoids and a
combination of sinusoids with amplitude up to 50 % of the
peak- to-peak of ECG amplitude. The model of power line
interference is provided as in (1). Fifty Hertz power line noise
is simulated using the MATLAB® [6]. The noise level
corresponds to the peak-to-peak amplitude of 0.15 mV. The
frequency of power line is 50 Hz.

N(t) =A×sin (2×π× f ×t) (1)

Where N(t) is the power line noise, A is the amplitude and f

is the frequency of power line.
Analog notch filter or band stop filter is mostly the first Fig. 2 Noise free ECG signal
choice for rejecting the specific frequency of the signal. The
notch filters were designed to suppress the 50 Hz power line

http://dx.doi.org/10.15242/ IJCCIE.E1113078 89
 Int'l Journal of Computing, Communications & Instrumentation Engg. (IJCCIE) Vol. 1, Issue 1 (2014) ISSN 2349-1469 EISSN 2349-1477

Fig. 3 ECG signal corrupted with noise sources

IV. RESULTS AND DISCUSSIONS

Five standard filter configurations have been tested for high
pass and notch filter implementations to suppress both
Baseline wandering noise and power line interference noise. Fig. 4 Magnitude response of the Elliptical High pass filter
Their performances are compared by computing the Signal-to-
Noise ratio (SNR) for each of them. The results of this
comparison are summarized in the Table II.
A 4th order Elliptic filter and a 7th order Chebyshev Type II
configurations are found to be better choice for the
implementation of high pass and Notch filters respectively.

TABLE II
COMPARISON OF PARAMETERS OF VARIOUS FILTERING APPROACHES
Highpass Filter Notch Filter
Filter Order: 4 Filter Order: 7
Input SNR= -11.6132 Input SNR= -8.2286
Filter Type SNR (dB) SNR (dB)

Butterworth 2.4946 8.9274

Chebyshev-I 3.3689 8.7764

Chebyshev-II 2.5501 8.9676 Fig. 5 ECG signal after high pass filtration

Elliptic 3.3846 8.9669

B. Power Line Interference Removal by Chebyshev II Notch
A. Baseline Wander Removal by Elliptic High Pass Filter Filter
Baseline wandering is eliminated by using an Elliptical high The frequency component at 50 Hz can be removed by
pass filter [7]. After we remove baseline wandering, the using a notch filter, in which a zero is placed on the unit circle
resulting ECG signal is more stationary and explicit than the at the location corresponding to 50Hz.
original signal.
The high pass elliptic filter is designed with order of 4 with V. CONCLUSION
the attenuation in pass band is 1 dB, the attenuation in the stop
Results of this investigation have shown that both Elliptic
band is 60dB and cut-off frequency 0.5Hz. Fig. 2-4 shows the
high pass filter and Chebyshev Type II notch filter methods are
responses of the elliptic high pass filter.
better choice for the successful removal of major components
From responses of the filter following observations are
of noise from the ECG signal. As the order of filtering system
made:
is increased, so the signal to noise ratio improves. On the other
 This filter provides equiripple behavior in the pass as hand, increasing the order of the filter increases the number of
well as stop band. components and hence more power dissipation. The proposed
 Pole-zero diagrams clearly shows that poles and zeros filter structure is based on satisfying both an acceptable signal
are within the circle so that the designed filter is stable. to noise ratio and real time analysis of the ECG signal.
Fig. 4 shows the magnitude response of the elliptic high
pass filter of the order 4 and cutoff frequency 0.5Hz. The ECG
signal after passing through the high pass filter is shown in the
Fig. 5. It can be seen that the signal now is free from baseline
wander noise.

http://dx.doi.org/10.15242/ IJCCIE.E1113078 90
 Int'l Journal of Computing, Communications & Instrumentation Engg. (IJCCIE) Vol. 1, Issue 1 (2014) ISSN 2349-1469 EISSN 2349-1477

Fig. 6 Magnitude and Phase response of the Chebyshev-II notch filter

Fig. 7 ECG signal after power line artifact filtration

REFERENCES [5] Aung Soe Khaing and Zaw Min Naing, “Quantitative Investigation of
Digital Filters in Electrocardiogram with Simulated Noises”,
[1] M. S. Chavan, R. A.Agarwala, M.D.Uplane, “Suppression of Baseline International Journal of Information and Electronics Engineering, Vol.
Wander and power line interference in ECG using Digital IIR Filter”, 1, No. 3, November 2011
International Journal of Circuits, Systems and Signal Processing, Issue [6] http://www.mathworks.in/
2, Volume 2, 2008 [7] M. K. Islam, A. N. M. M. Haque, G. Tangim, T. Ahammad, and M. R.
[2] http://www.physionet.org/cgi-bin/atm/ATM – “MIT-BIH Arrhythmia H. Khondokar, “Study and Analysis of ECG Signal Using MATLAB &
Database”. LABVIEW as Effective Tools”, International Journal of Computer and
[3] Geng Yang, Ying Cao, Jian Chen, Hannu Tenhunen, Li-Rong Zheng, Electrical Engineering, Vol. 4, No. 3, June 2012.
“An Active-Cable Connected ECG Monitoring System for Ubiquitous
Healthcare”, IEEE International Conference on Convergence and
Hybrid Information Technology, 2008.
[4] D. Jeyarani, T. Jaya Singh, “Analysis of Noise Reduction Techniques on
QRS ECG Waveform - by Applying Different Filters”, IEEE conference
on Recent Advances in Space Technology Services and Climate Change
(RSTSCC), Chennai, 2010

http://dx.doi.org/10.15242/ IJCCIE.E1113078 91