1

Detection of QRS complex in ECG signal using

Multiresolution Wavelet and Thresholding Method

Soroor Behbahani 1 Nader Jafarnia Dabanloo 2

1- Biomedical Engineering Department, Islamic Azad University, Science and Research Branch, Tehran, Iran
[email protected]
2- Biomedical Engineering Department, Islamic Azad University, Science and Research Branch, Tehran, Iran.
[email protected]

Abstract
The electrocardiogram (ECG) is a diagnostic tool that
measures and records the electrical activity of the
heart in exquisite detail. Interpretation of these details
allows diagnosis of a wide range of heart conditions.
Automatic extraction of time plane features is
important for cardiac disease diagnosis. This paper
presents a multi-resolution wavelet transform based
system for detection and evaluation of QRS complex.
In first step of this paper we use the selective confident
method to find the QRS complex, at next step we use a
threshold method to find the QRS complex and finally
we apply the composition of first step algorithm and
thresholding method which shows robust ability of
finding QRS compared to other methods. Achieved
overall accuracy of R and QRS detection for
only
4
d scale without threshold is 84.48%, the
composition of
5 4 3
, , d d d without threshold 93.23%,
only
4
d with threshold 90%, and
5 4 3
, , d d d with
threshold 98.2%.

Keywords: ECG signal, QRS waves, Feature
Extraction, threshold method, Db6 wavelets.

1. Introduction
The electrocardiogram is an important tool for providing
information about functional status of the heart. Analysis
of ECG is of great importance in the detection of cardiac
diseases which are the main cause of mortality in many
countries.
The Q, R and S wave of ECG signal occur in rapid
succession, do not all appear in all leads and reflect a
single event so are thus normally considered as a whole
complex. A Q wave is any downward deflection after the
P-wave. An R-wave is an upward deflection and the S
wave is any downward deflection after the R-wave.
The QRS complex of the electrocardiographic signal has
the normal duration from 0.06s to 0.1s and provides
information about the heart rate, the conduction velocity,
the condition of tissues within the heart and various




abnormalities. Its shape, duration and time of occurrence
provide valuable information about the current state of the
heart. Because of its specific shape, the QRS complex
serves as an entry point for almost all automated ECG
analysis algorithms and detection of the QRS complex is
the most important task in automatic ECG signal analysis
[1].
Ventricles contain more muscle mass than the atria,
therefore the QRS complex is considerably larger than the
P wave. The QRS complex is often used to determine the
axis of the electrocardiogram .The atrial repolarization
wave, which resembles an inverse P wave, is buried
inside the QRS wave. The atrial repolarization wave is
obscured by the QRS because it is far smaller in
magnitude. The duration, amplitude, and morphology of
the QRS complex is useful in diagnosing cardiac
arrhythmias, conduction abnormalities, ventricular
hypertrophy, myocardial infarction, electrolyte
derangements, and other disease states.
The QRS detection is not a simple task, due to the varying
morphologies of normal and abnormal complexes and
because the ECG signal experiences different types of
disturbances with complex origin [2].
There are several methods for detection of significant
points of ECG. Some of these methods are, low pass
differentiation (LPD), Multichannel ART-based neural
network (MART), wavelet transform (WT), and etc.
In signal processing, there are a number of
different functions that can perform on the
signal in order to translate it into different forms that
are more suitable for different applications. The most
popular function is the Fourier transform that converts a
signal from time versus amplitude to frequency versus
amplitude. This transform is useful for many applications,
but it is not based in time. To combat this problem,
mathematicians came up with the short term Fourier
transform which can convert a signal to frequency versus
time. Unfortunately, this transform also has its
shortcomings mostly that it cannot get decent resolutions
for both high and low frequencies at the same time.
So to converted and manipulated the signal while keeping
resolution across the entire signal and still being based in
time, wavelets come into play. Wavelets are finite
2
windows through which the signal can be viewed. In
order to move the window about the length of the signal,
the wavelets can be translated about time in addition to
being compressed and widened.
Using DWT in feature extraction may lead to an optimal
frequency resolution in all frequency ranges as it has a
varying window size, broad at lower frequencies, and
narrow at higher frequencies. The DWT characterization
will deliver the stable features to the morphology
variations of the ECG waveforms [3].
This paper presents a multi-resolution wavelet transform
based system for detection and evaluation of QRS
complex. In first step of this paper we use the represented
algorithm in [2], and fined the QRS complex, at next step
we use a threshold method to find the QRS complex and
finally we apply the composition of first step algorithm
and thresholding method which shows robust ability of
finding QRS in compared to other methods. The
performance of the system is validated using original 12
lead ECG recording collected from the Physionet PTB
diagnostic database [8].

2. Discrete Wavelet Transform (DWT)
A wavelet is a waveform of effectively limited duration
that has an average value of zero. Similar to Fourier series
analysis, where sinusoids are chosen as the basis function,
wavelet analysis is also based on a decomposition of a
signal using an orthonormal (typically, although not
necessarily) family of basic functions. Unlike a sine
wave, a wavelet has its energy concentrated in time.
Sinusoids are useful in analyzing periodic and time-
invariant phenomena, while wavelets are well suited for
the analysis of transient, time-varying signals, thus well
suited for ECG signals. Basically wavelet transform is the
convolution operation of the subject signal ) (t f and the
wavelet function ) (t . The discrete wavelet transform is
expressed as,
}

= dt t t f X
k j k j
) ( ) (
, ,
(1)
The approximation coefficient of the signal ) (t f is
represented as,
}

= dt t t f A
k j k j
) ( ) (
, ,
| (2)
Where ) (t | is scaling function, j and k are scale and
location respectively. For a range of scale n, the original
signal ) (t f under discrete wavelet transform can be
represented as,

=
+ =
n
j
j n
t d t f t f
1
) ( ) ( ) ( (3)
Where ) (t f
n
is mean signal approximation and is given
by,
) ( ) (
, ,
t A t f
k n k n n
| = (4)
and ) (t d
j
is detail signal approximation in scale j .
Thus given an approximation of a signal using
translations of a mother wavelet up to some chosen scale,
a better approximation can be achieved by using
expansion signals with half the width and half as wide
translation steps. The wavelet transform as such
decomposes a signal into two sub signals detail signal
and approximation signal. Detail signal contains the upper
half of the frequency components and approximation
signal contains the lower half.
The decomposition can be further repeated on the
approximation signal in order to g t the second detail and
approximation signal. Thus in discrete wavelet domain,
multi-resolution analysis can be performed.

2.1. Wavelet Selection
The large number of known wavelet families and
functions provides a rich space in which to search for a
wavelet which will very efficiently represent a signal of
interest in a large variety of applications. Wavelet
families include Biorthogonal, Coiflet, Harr, Symmlet,
Daubechies wavelets, etc. There is no absolute way to
choose a certain wavelet. The choice of the wavelet
function depends on the application. To obtain the
wavelet analysis, we used the Matlab program, which
contains a very good wavelet toolbox. First the
considered ECG signal was decomposed using db6
wavelet of the order of 1-8 has been evaluated (Figure2).

3. Peaks Detection
In order to detect the peaks, specific details of the signal
were selected. The method developed by Li et al [4]
involves preprocessing of initial QRS beats to select
modulus maxima greater than threshold and post
processing to remove the unrelated noisy peaks appearing
as R-waves. The method uses four thresholds for the
detection of modulus maxima at four different scales.
Mahmood abadi[5] suggest the selection of details d3-d5
for R-Wave detection, whereas S.C.Saxena et al [6]
employs detail signal d4 for detection of QRS peak.
Awadhesh Pachauri, and Manabendra Bhuyan[7]
represent a method without requirement of preprocessing
and the selection of detail signal d4 has been justified by
energy, frequency and correlation analysis. Furthermore,
they calculate total number of R waves using hard
thresholding.
In this paper we use we apply the composition of use the
selective confident method [2] and thresholding method
3
which shows robust ability of finding QRS in compared
to other methods.










































Figure1: Wavelet coefficients for scale levels 18


3.1. R peak Detection
R peak is the largest amplitude point which is greater than
threshold points are located in the wave.
It can be seen in figure1 that small scales (1-4), represent
the high frequency components and large scales represent
the low frequency components of sample ECG signal.
Eighth level reconstruction coefficients represent high
frequency contents of the ECG waveform which in most
of the cases appear to be high frequency noises.
It is clear that most energy of the QRS complex is
concentrated at decomposition level 3, 4 and 5. Thus, d3,
d4 and d5 coefficients are identified for the detection of
QRS complex. The plot of the reconstructed wave
comprising of, d3, d4 and d5 is shown in Figure 2.
5 4 3 1
d d d e + + = (5)
From Figure 2, it is clear that although the QRS region is
properly captured but it is difficult to identify R peak due
to its oscillatory nature. So, a function
2
e is defined as
[2],
n
d d d
e
2
) (
5 3 4
2
+
= (6)
where n is the level of decomposition. Then the modulus
of
2 1
e e is taken. The corresponding plot is shown in
figure 4 along with the original wave. The accuracy of the




entire feature extraction work mainly depends upon the
identification accuracy of R peak.

3.2. Thresholding Method
The selected detail coefficient
4
d is used to perform the
detection of R-wave. For this purpose a practical lower
limit is applied to remove unrelated peaks appearing due
to noise.
There are several thresholding methods known. Here we
apply hard thresholding in which the samples below a
predetermined threshold are set to zero. The threshold is
selected as 15% of maximum value of
4
d and applied as
follows [4]:
0 )) ( (
)) ( (
) max( 15 %
4
4
4
=
=
=
i d
threshold i d if
d threshold

In some researches the detail coefficient
4
d has been
chosen to detect R wave based on Energy, frequency and
correlation Analysis [4]. In this paper we compared the
results of R detection based on only
4
d , and composition
of
5 4 3
, , d d d , also we apply both of these cases with and
without using threshold method. The results are shown in
figure 2-5 respectively.




0 1000 2000 3000 4000 5000
-10
-5
0
5
x 10
-4
d1
0 1000 2000 3000 4000 5000
-5
0
5
x 10
-3
d2
0 1000 2000 3000 4000 5000
-0.05
0
0.05
d3
0 1000 2000 3000 4000 5000
-0.5
0
0.5
d4
0 1000 2000 3000 4000 5000
-0.5
0
0.5
d5
0 1000 2000 3000 4000 5000
-0.5
0
0.5
d6
0 1000 2000 3000 4000 5000
-0.2
0
0.2
d7
0 1000 2000 3000 4000 5000
-0.1
0
0.1
d8
4












Figure 2: Original signal and plot of
5 4 3 1
d d d e + + =

3.3. Q and S Detection
A Q and S peak occurs about the R peak with in
0.1second. Calculating the distance from zero point or
close zero left side of R peak within the threshold limit
denotes Q peak.
Once the R peak is detected, the Q and S points are to be
identified to detect the complete QRS complex. Generally
the Q and S waves have high frequency and low
amplitude and their energies are mainly at small scale. So,
decomposition coefficients from d2 to d5 are used.
Q and S points are the points of inflexion in either side of
the R peak. So the first zero slope points on either side of
the R peak will represent the Q and S point. Figure 5
shows the detected QRS complex.First two zero slope
points on either side of R peak (as detected earlier) are
identified as Q and S points respectively.

3.4. QRS Detection with baseline drift
In order to evaluate the represented algorithm of finding
QRS complex and R wave, we apply this algorithm on an
ECG signal with baseline drift. Since there was not an
available ECG signal with baseline drift, we use a
manipulate signal which made from an ECG signal added
to a random sinuous and cosines wave. Then the
algorithm was applied on manipulate signal. The results
of QRS and R wave detection are shown in figure 7 and 8
respectively.

4. Results
In the present work, physionet PTB diagnostic database is
used to detect the QRS complex. As discussed in the
previous section, simulation and testing has been carried
out, the results of which are presented in Figures 1-8. For
validation of the algorithm, 50 databases were checked.
Achieved overall accuracy of R and QRS detection for
only
4
d scale without threshold is 84.48%, the
composition of
5 4 3
, , d d d without threshold 93.23%,
only
4
d with threshold 90%, and
5 4 3
, , d d d with
threshold 98.2%.















Figure 3: Original signal and plot of modulus of
2 1
e e
















Figure 4: QRS detection on ECG signal

5. Conclusion
In this paper, an algorithm based on Wavelet Transform is
presented for the detection of QRS complex of ECG
signal. Wavelet Decomposition of ECG wave up to level
8 using orthogonal Daubechis 6 wavelet generates 8
scales of detail coefficients. Smaller scales correspond to
high frequency components and higher scales correspond
to low frequency components of the signals.
The multi-resolution wavelet transform based system with
different wavelet scales used to find the QRS complex,
and finally the composition of selective confident and
thresholding method shows robust ability of finding QRS
compared to other methods.





0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
-0.5
0
0.5
1
original signal
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
-0.5
0
0.5
1
e1
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
-0.5
0
0.5
1
original signal
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
-2
0
2
4
x 10
-4
e1 x e2
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
-0.2
0
0.2
0.4
0.6
0.8
1
1.2
QRS
5
0 100 200 300 400 500 600 700 800 900 1000
0
100
200
300
400
500
R Detection With Thresholding(d4)
0 100 200 300 400 500 600 700 800 900 1000
0
200
400
600
800
R Detection With Thresholding(e1*e2)















Figure 5: R detection with and without threshold based on
5 4 3
, , d d d















Figure 7: QRS detection on ECG signal with baseline drift

Figure 6: Comparison of R detection with threshold based on
5 4 3
, , d d d and
4
d
















Figure 8: R wave detection on ECG signal with baseline drift

From the result, it can be seen that this method is having a
Comparatively high sensitivity, so it can be used as a
useful and robust tool for automatic and on line disease
classification and biometric recognition. Since the
application of wavelet transformation in electro
cardiology is relatively new field of research, many
methodological aspects (Choice of the mother wavelet,
values of the scale parameters) of the wavelet technique
will require further investigations in order to improve the
clinical usefulness of this novel signal processing
technique. Simultaneously diagnostic and prognostic
significance of wavelet techniques in various fields of
electro cardiology needs to be established in large clinical
studies. But some of the ECG waveform may show very
erratic nature due to electrode contact noise or some
complicated cardiac abnormalities. The algorithm is not
tested with them because of lack of availability of that
special kind of database.


6. References
[1] C. Li, C. Zheng, C. Tai, Detection of ECG characteristic points by
wavelet transforms, IEEE Transactions on Biomedical Engineering 42(1)
(1995) 2128.
[2] Saurabh Pal, Madhuchhanda Mitra, Detection of ECG characteristic
points using Multiresolution WaveletAnalysis based Selective
Coefficient Method, Elsevier, Measurement 43 (2010), 255261.
[3] S. Karpagachelvi, Dr. M.Arthanari, Prof. & Head, M. Sivakumar,
QRS Wave Detection Using Multiresolution Analysis, Vol. 10 Issue 5
Ver. 1.0, (2010), 39-42.
[4] Cuiwei Li, Chongxun Zheng, and Changfeng Tai, Detection of ECG
Characteristic points using Wavelet Transforms, IEEE Trans. Biomed.
Eng, Vol. 42, No. 1, (1995).
[5] S. Z. Mahmoodabadi, A .Ahmadian , M. D.Abolhasani, M. Eslami, J.
H.Bidgoli, ECG Feature Extraction Based on Multiresolution Wavelet
Transform, Proceedings of the 2005 IEEE, Engineering in Medicine
and Biology 27th Annual Conference Shanghai, China, September 1-
4,(2005).
[6] S.C. Saxena, V. Kumar and S.T. Hande, QRS Detection using New
Wavelets, Journal of Medical Engineering & Technology, Volume 26,
November1, pages 7-15, (2002).
[7]Awadhesh Pachauri, and Manabendra Bhuyan, Robust Detection of
R-Wave Using Wavelet Technique, World Academy of Science,
Engineering and Technology 56, (2009).
[8] http://www.physionet.org/mitdb

0 100 200 300 400 500 600 700 800 900 1000
-200
0
200
400
600
800
R Detection Without Thresholding(e1*e2)
0 100 200 300 400 500 600 700 800 900 1000
0
200
400
600
800
R Detection With Thresholding(e1*e2)
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
2
4
6
8
Manipulated signal
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
-2
0
2
4
x 10
-4
e1 x e2
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
2
4
6
8
Manipulated signal
0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000
-0.5
0
0.5
1
e1